{"id":105388,"date":"2026-11-01T20:41:22","date_gmt":"2026-11-01T18:41:22","guid":{"rendered":"https:\/\/twodots.gr\/?p=105388"},"modified":"2026-11-01T20:41:22","modified_gmt":"2026-11-01T18:41:22","slug":"reasoning-tax-llms-perissoteri-skepsi-kostos","status":"publish","type":"post","link":"https:\/\/twodots.gr\/hr\/reasoning-tax-llms-perissoteri-skepsi-kostos\/","title":{"rendered":"\u00abPorez na razmi\u0161ljanje\u00bb kod LLM-ova: kada je vi\u0161e razmi\u0161ljanja vrijedno tro\u0161ka?"},"content":{"rendered":"<p><strong>\u03a4\u03bf \u00abreasoning tax\u00bb \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf \u03c0\u03c1\u03cc\u03c3\u03b8\u03b5\u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03c3\u03b5 generated tokens, latency \u03ba\u03b1\u03b9 \u03c5\u03c0\u03bf\u03b4\u03bf\u03bc\u03ae \u03c0\u03bf\u03c5 \u03c0\u03bb\u03b7\u03c1\u03ce\u03bd\u03b5\u03b9 \u03bc\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03af\u03c1\u03b7\u03c3\u03b7 \u03cc\u03c4\u03b1\u03bd \u03ad\u03bd\u03b1 LLM \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03b5\u03ba\u03c4\u03b5\u03c4\u03b1\u03bc\u03ad\u03bd\u03b7 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03ae \u03b5\u03c0\u03b5\u03be\u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1. \u0391\u03be\u03af\u03b6\u03b5\u03b9 \u03bc\u03cc\u03bd\u03bf \u03cc\u03c4\u03b1\u03bd \u03c4\u03bf \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03b9\u03bc\u03bf \u03ba\u03ad\u03c1\u03b4\u03bf\u03c2 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 \u03c3\u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf task \u03c5\u03c0\u03b5\u03c1\u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03b5\u03b9 \u03b1\u03c5\u03c4\u03cc \u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2.<\/strong><\/p>\n<p>\u039f\u03b9 Sachin Gopal Wani, Ajay Dholakia \u03ba\u03b1\u03b9 David Ellison \u03b5\u03be\u03b5\u03c4\u03ac\u03b6\u03bf\u03c5\u03bd \u03b1\u03c5\u03c4\u03cc \u03c4\u03bf \u03b5\u03c1\u03ce\u03c4\u03b7\u03bc\u03b1 \u03c3\u03c4\u03b7\u03bd \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 <em>The Reasoning Tax: Token Economics of LLM Reasoning Across Task Types and Deployment Contexts<\/em>. \u0391\u03bd\u03b1\u03bb\u03cd\u03bf\u03c5\u03bd 151 model-benchmark runs \u03c3\u03b5 \u03b5\u03c0\u03c4\u03ac benchmarks \u03ba\u03b1\u03b9 27 \u03b4\u03b9\u03b1\u03bc\u03bf\u03c1\u03c6\u03ce\u03c3\u03b5\u03b9\u03c2 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03c9\u03bd \u03b1\u03c0\u03cc \u03bf\u03ba\u03c4\u03ce \u03bf\u03b9\u03ba\u03bf\u03b3\u03ad\u03bd\u03b5\u03b9\u03b5\u03c2. \u03a4\u03bf \u03b2\u03b1\u03c3\u03b9\u03ba\u03cc \u03c3\u03c5\u03bc\u03c0\u03ad\u03c1\u03b1\u03c3\u03bc\u03b1 \u03b3\u03b9\u03b1 \u03cc\u03c3\u03bf\u03c5\u03c2 \u03c3\u03c7\u03b5\u03b4\u03b9\u03ac\u03b6\u03bf\u03c5\u03bd <a href=\"https:\/\/twodots.gr\/ibm-granite-42-reasoning-ai-agents-enterprise\/\">reasoning \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03c3\u03b5 enterprise AI workflows<\/a> \u03b5\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03b7 \u03b4\u03bf\u03bc\u03ae \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2 \u03c0\u03c1\u03bf\u03b2\u03bb\u03ad\u03c0\u03b5\u03b9 \u03c4\u03b7\u03bd \u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ae \u03b1\u03be\u03af\u03b1 \u03c4\u03bf\u03c5 reasoning \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b1 \u03b1\u03c0\u03cc \u03c4\u03b7 \u03b3\u03b5\u03bd\u03b9\u03ba\u03ae \u03b5\u03c4\u03b9\u03ba\u03ad\u03c4\u03b1 \u00ab\u03b5\u03cd\u03ba\u03bf\u03bb\u03bf\u00bb \u03ae \u00ab\u03b4\u03cd\u03c3\u03ba\u03bf\u03bb\u03bf\u00bb.<\/p>\n<div class=\"td-article-note\"><strong>\u0391\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u03c3\u03b5 \u03bc\u03af\u03b1 \u03c0\u03c1\u03cc\u03c4\u03b1\u03c3\u03b7:<\/strong> \u03b5\u03bd\u03b5\u03c1\u03b3\u03bf\u03c0\u03bf\u03b9\u03ae\u03c3\u03c4\u03b5 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf reasoning \u03bc\u03cc\u03bd\u03bf \u03c3\u03c4\u03b1 \u03b2\u03ae\u03bc\u03b1\u03c4\u03b1 \u03cc\u03c0\u03bf\u03c5 \u03ad\u03bd\u03b1 \u03b1\u03bd\u03c4\u03b9\u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03b5\u03c5\u03c4\u03b9\u03ba\u03cc validation set \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03b7 \u03c0\u03c1\u03cc\u03c3\u03b8\u03b5\u03c4\u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03b4\u03b9\u03ba\u03b1\u03b9\u03bf\u03bb\u03bf\u03b3\u03b5\u03af \u03c4\u03b1 tokens, \u03c4\u03bf\u03bd \u03c7\u03c1\u03cc\u03bd\u03bf \u03b1\u03c0\u03cc\u03ba\u03c1\u03b9\u03c3\u03b7\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03bf \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03af\u03b1\u03c2.<\/div>\n<div class=\"td-article-toc\">\n<div class=\"td-toc-title\">Sadr\u017eaj<\/div>\n<ul>\n<li><a href=\"#oikonomia-kathe-token\">\u0391\u03c0\u03cc \u03c4\u03bf accuracy benchmark \u03c3\u03c4\u03b7\u03bd \u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03af\u03b1 \u03c4\u03bf\u03c5 \u03ba\u03ac\u03b8\u03b5 token<\/a><\/li>\n<li><a href=\"#token-economy-score\">\u03a4\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c4\u03bf Token Economy Score<\/a><\/li>\n<li><a href=\"#tes-delta-tes-a\">\u0394\u03cd\u03bf \u03c4\u03c1\u03cc\u03c0\u03bf\u03b9 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7\u03c2: TES-\u0394 \u03ba\u03b1\u03b9 TES-A<\/a><\/li>\n<li><a href=\"#domi-ergasias\">\u0397 \u03b4\u03bf\u03bc\u03ae \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf \u03b1\u03c0\u03cc \u03c4\u03b7 \u03b4\u03c5\u03c3\u03ba\u03bf\u03bb\u03af\u03b1<\/a><\/li>\n<li><a href=\"#maximum-reasoning\">\u03a4\u03bf maximum reasoning \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03bf\u03c5\u03bc\u03c0\u03af \u00ab\u03bc\u03ad\u03b3\u03b9\u03c3\u03c4\u03b7\u03c2 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2\u00bb<\/a><\/li>\n<li><a href=\"#reasoning-cost-share\">Reasoning Cost Share: \u03c0\u03bf\u03cd \u03c0\u03b7\u03b3\u03b1\u03af\u03bd\u03b5\u03b9 \u03bf \u03bb\u03bf\u03b3\u03b1\u03c1\u03b9\u03b1\u03c3\u03bc\u03cc\u03c2<\/a><\/li>\n<li><a href=\"#cloud-on-premises\">Cloud \u03ae on-premises: \u03c4\u03bf \u03af\u03b4\u03b9\u03bf TES, \u03ac\u03bb\u03bb\u03b7 \u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ae \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc\u03c4\u03b7\u03c4\u03b1<\/a><\/li>\n<li><a href=\"#tessera-tetartimoria\">\u03a4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 \u03c4\u03b5\u03c4\u03b1\u03c1\u03c4\u03b7\u03bc\u03cc\u03c1\u03b9\u03b1 \u03b3\u03b9\u03b1 \u03c4\u03b7\u03bd \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5<\/a><\/li>\n<li><a href=\"#ai-agents-workflows\">\u03a4\u03b9 \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03c3\u03b5 AI agents \u03ba\u03b1\u03b9 \u03c0\u03bf\u03bb\u03c5\u03b2\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac workflows<\/a><\/li>\n<li><a href=\"#plano-axiologisis\">\u0388\u03bd\u03b1 \u03c0\u03bb\u03ac\u03bd\u03bf \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7\u03c2 \u03b3\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03ae\u03c3\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 e-commerce teams<\/a><\/li>\n<li><a href=\"#anagnosi-arithmon\">\u03a0\u03ce\u03c2 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03b2\u03ac\u03c3\u03b5\u03c4\u03b5 \u03c4\u03b1 \u03bd\u03bf\u03cd\u03bc\u03b5\u03c1\u03b1 \u03c7\u03c9\u03c1\u03af\u03c2 \u03bd\u03b1 \u03c4\u03b1 \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c8\u03b5\u03c4\u03b5 \u03c3\u03b5 \u03c5\u03c0\u03cc\u03c3\u03c7\u03b5\u03c3\u03b7<\/a><\/li>\n<li><a href=\"#periorismoi\">\u03a0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03af \u03c0\u03bf\u03c5 \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03ba\u03c1\u03c5\u03c6\u03c4\u03bf\u03cd\u03bd<\/a><\/li>\n<li><a href=\"#symperasma\">\u0397 \u03bf\u03c5\u03c3\u03af\u03b1 \u03b3\u03b9\u03b1 \u03bc\u03b9\u03b1 \u03b4\u03b9\u03bf\u03af\u03ba\u03b7\u03c3\u03b7 \u03c0\u03bf\u03c5 \u03b5\u03c0\u03b5\u03bd\u03b4\u03cd\u03b5\u03b9 \u03c3\u03b5 AI<\/a><\/li>\n<\/ul>\n<\/div>\n<h2 id=\"oikonomia-kathe-token\">\u0391\u03c0\u03cc \u03c4\u03bf accuracy benchmark \u03c3\u03c4\u03b7\u03bd \u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03af\u03b1 \u03c4\u03bf\u03c5 \u03ba\u03ac\u03b8\u03b5 token<\/h2>\n<p>\u03a4\u03b1 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b1 leaderboards \u03b1\u03c0\u03b1\u03bd\u03c4\u03bf\u03cd\u03bd \u03c3\u03c4\u03bf \u00ab\u03c0\u03bf\u03b9\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03ad\u03b4\u03c9\u03c3\u03b5 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b5\u03c2 \u03c3\u03c9\u03c3\u03c4\u03ad\u03c2 \u03b1\u03c0\u03b1\u03bd\u03c4\u03ae\u03c3\u03b5\u03b9\u03c2;\u00bb. \u0394\u03b5\u03bd \u03b1\u03c0\u03b1\u03bd\u03c4\u03bf\u03cd\u03bd \u03cc\u03bc\u03c9\u03c2 \u03c3\u03c4\u03bf \u00ab\u03c0\u03cc\u03c3\u03bf \u03b5\u03c0\u03b9\u03c0\u03bb\u03ad\u03bf\u03bd \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03c4\u03b9\u03ba\u03cc \u03ad\u03c1\u03b3\u03bf \u03c7\u03c1\u03b5\u03b9\u03ac\u03c3\u03c4\u03b7\u03ba\u03b5 \u03b3\u03b9\u03b1 \u03b1\u03c5\u03c4\u03ae \u03c4\u03b7 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac;\u00bb. \u0388\u03bd\u03b1 reasoning model \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03ba\u03b5\u03c1\u03b4\u03af\u03c3\u03b5\u03b9 \u03bb\u03af\u03b3\u03b5\u03c2 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b5\u03c2 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2, \u03c0\u03b1\u03c1\u03ac\u03b3\u03bf\u03bd\u03c4\u03b1\u03c2 \u03c0\u03bf\u03bb\u03bb\u03b1\u03c0\u03bb\u03ac\u03c3\u03b9\u03b1 tokens \u03c3\u03b5 \u03c3\u03c7\u03ad\u03c3\u03b7 \u03bc\u03b5 \u03ad\u03bd\u03b1 instruction-following baseline. \u0393\u03b9\u03b1 \u03bc\u03b9\u03b1 \u03bc\u03b5\u03bc\u03bf\u03bd\u03c9\u03bc\u03ad\u03bd\u03b7 \u03b4\u03bf\u03ba\u03b9\u03bc\u03ae \u03b1\u03c5\u03c4\u03cc \u03af\u03c3\u03c9\u03c2 \u03c6\u03b1\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 \u03b1\u03bc\u03b5\u03bb\u03b7\u03c4\u03ad\u03bf. \u03a3\u03b5 \u03c7\u03b9\u03bb\u03b9\u03ac\u03b4\u03b5\u03c2 \u03ba\u03bb\u03ae\u03c3\u03b5\u03b9\u03c2, \u03c3\u03b5 \u03ad\u03bd\u03b1\u03bd agent \u03bc\u03b5 \u03c0\u03bf\u03bb\u03bb\u03ac \u03b2\u03ae\u03bc\u03b1\u03c4\u03b1 \u03ae \u03c3\u03b5 \u03ad\u03bd\u03b1 customer-support workflow, \u03b7 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03c3\u03c5\u03c3\u03c3\u03c9\u03c1\u03b5\u03cd\u03b5\u03c4\u03b1\u03b9.<\/p>\n<p>\u0397 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7 \u03b1\u03c0\u03bf\u03bc\u03bf\u03bd\u03ce\u03bd\u03b5\u03b9 \u03c4\u03b1 generated tokens, \u03b4\u03b7\u03bb\u03b1\u03b4\u03ae reasoning tokens \u03ba\u03b1\u03b9 tokens \u03c4\u03b7\u03c2 \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae\u03c2 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7\u03c2. \u03a4\u03b1 input tokens \u03bc\u03ad\u03bd\u03bf\u03c5\u03bd \u03b5\u03ba\u03c4\u03cc\u03c2 \u03c4\u03bf\u03c5 \u03b2\u03b1\u03c3\u03b9\u03ba\u03bf\u03cd metric, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03b5\u03c0\u03b7\u03c1\u03b5\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03ba\u03c5\u03c1\u03af\u03c9\u03c2 \u03b1\u03c0\u03cc \u03c4\u03bf prompt \u03ba\u03b1\u03b9 \u03c4\u03bf evaluation harness. \u0388\u03c4\u03c3\u03b9, \u03b7 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7 \u03b5\u03c3\u03c4\u03b9\u03ac\u03b6\u03b5\u03b9 \u03c3\u03c4\u03bf \u03b5\u03c0\u03b9\u03c0\u03bb\u03ad\u03bf\u03bd \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03c0\u03bf\u03c5 \u03c0\u03c1\u03bf\u03ba\u03b1\u03bb\u03b5\u03af \u03b7 \u03af\u03b4\u03b9\u03b1 \u03b7 \u03c3\u03c5\u03bc\u03c0\u03b5\u03c1\u03b9\u03c6\u03bf\u03c1\u03ac reasoning. \u0397 \u03c0\u03c1\u03bf\u03c3\u03ad\u03b3\u03b3\u03b9\u03c3\u03b7 \u03b4\u03b5\u03bd \u03b1\u03ba\u03c5\u03c1\u03ce\u03bd\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u0387 \u03c4\u03b7 \u03c3\u03c5\u03bd\u03b4\u03ad\u03b5\u03b9 \u03bc\u03b5 \u03c4\u03bf \u03c4\u03af\u03bc\u03b7\u03bc\u03b1 \u03c0\u03bf\u03c5 \u03b1\u03c0\u03b1\u03b9\u03c4\u03b5\u03af\u03c4\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03bd\u03b1 \u03b1\u03c0\u03bf\u03ba\u03c4\u03b7\u03b8\u03b5\u03af.<\/p>\n<h2 id=\"token-economy-score\">\u03a4\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c4\u03bf Token Economy Score<\/h2>\n<p>\u03a4\u03bf Token Economy Score, \u03ae TES, \u03bc\u03b5\u03c4\u03c1\u03ac \u03c4\u03b7 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2 \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03b5 \u03ad\u03bd\u03b1 reasoning model \u03ba\u03b1\u03b9 \u03ad\u03bd\u03b1 non-reasoning baseline \u03ba\u03b1\u03b9 \u03c4\u03b7 \u03b4\u03b9\u03b1\u03b9\u03c1\u03b5\u03af \u03bc\u03b5 \u03c4\u03bf\u03bd \u03c0\u03bf\u03bb\u03bb\u03b1\u03c0\u03bb\u03b1\u03c3\u03b9\u03b1\u03c3\u03c4\u03ae generated tokens. \u0391\u03bd, \u03b3\u03b9\u03b1 \u03c0\u03b1\u03c1\u03ac\u03b4\u03b5\u03b9\u03b3\u03bc\u03b1, \u03c4\u03bf reasoning \u03c0\u03c1\u03bf\u03c3\u03b8\u03ad\u03c4\u03b5\u03b9 \u03c0\u03ad\u03bd\u03c4\u03b5 \u03c0\u03bf\u03c3\u03bf\u03c3\u03c4\u03b9\u03b1\u03af\u03b5\u03c2 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b5\u03c2 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2 \u03bc\u03b5 \u03c0\u03b5\u03bd\u03c4\u03b1\u03c0\u03bb\u03ac\u03c3\u03b9\u03b1 \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae tokens, \u03c4\u03bf TES \u03b5\u03af\u03bd\u03b1\u03b9 1. \u039f\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03bf\u03cd\u03bd \u03b1\u03c5\u03c4\u03ae \u03c4\u03b7\u03bd \u03c4\u03b9\u03bc\u03ae \u03c9\u03c2 \u03c3\u03c5\u03bd\u03b5\u03c0\u03ae \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03cc \u03ba\u03b1\u03bd\u03cc\u03bd\u03b1 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7\u03c2, \u03cc\u03c7\u03b9 \u03c9\u03c2 \u03ba\u03b1\u03b8\u03bf\u03bb\u03b9\u03ba\u03cc \u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03cc \u03bd\u03cc\u03bc\u03bf.<\/p>\n<p>TES \u03c0\u03ac\u03bd\u03c9 \u03b1\u03c0\u03cc 1 \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03c4\u03bf \u03ba\u03ad\u03c1\u03b4\u03bf\u03c2 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2 \u03c5\u03c0\u03b5\u03c1\u03b2\u03b1\u03af\u03bd\u03b5\u03b9 \u03c4\u03bf\u03bd token multiplier \u03ba\u03b1\u03b9 \u03c7\u03b1\u03c1\u03b1\u03ba\u03c4\u03b7\u03c1\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c5\u03c8\u03b7\u03bb\u03ac \u03b1\u03c0\u03bf\u03b4\u03bf\u03c4\u03b9\u03ba\u03cc. \u03a4\u03b9\u03bc\u03ae \u03b1\u03c0\u03cc 0 \u03ad\u03c9\u03c2 1 \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03b8\u03b5\u03c4\u03b9\u03ba\u03ae \u03b1\u03bb\u03bb\u03ac \u03bf\u03c1\u03b9\u03b1\u03ba\u03ae \u03b1\u03c0\u03cc\u03b4\u03bf\u03c3\u03b7. \u039c\u03b7\u03b4\u03b5\u03bd\u03b9\u03ba\u03cc \u03ae \u03b1\u03c1\u03bd\u03b7\u03c4\u03b9\u03ba\u03cc TES \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03c4\u03bf reasoning \u03b9\u03c3\u03bf\u03c6\u03b1\u03c1\u03af\u03b6\u03b5\u03b9 \u03ae \u03c5\u03c0\u03bf\u03bb\u03b5\u03af\u03c0\u03b5\u03c4\u03b1\u03b9 \u03c4\u03bf\u03c5 baseline, \u03c0\u03b1\u03c1\u03cc\u03c4\u03b9 \u03ba\u03b1\u03c4\u03b1\u03bd\u03b1\u03bb\u03ce\u03bd\u03b5\u03b9 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b1 tokens. \u0397 \u03c4\u03b5\u03bb\u03b5\u03c5\u03c4\u03b1\u03af\u03b1 \u03c0\u03b5\u03c1\u03af\u03c0\u03c4\u03c9\u03c3\u03b7 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b8\u03b5\u03c9\u03c1\u03b7\u03c4\u03b9\u03ba\u03ae: \u03b5\u03bc\u03c6\u03b1\u03bd\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03cc\u03c4\u03b1\u03bd \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c5\u03c0\u03b5\u03c1\u03b1\u03bd\u03b1\u03bb\u03cd\u03b5\u03b9, \u03b5\u03b9\u03c3\u03ac\u03b3\u03b5\u03b9 \u03ac\u03c3\u03c7\u03b5\u03c4\u03b5\u03c2 \u03b5\u03bd\u03b1\u03bb\u03bb\u03b1\u03ba\u03c4\u03b9\u03ba\u03ad\u03c2 \u03ae \u03c7\u03ac\u03bd\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03c1\u03c7\u03b9\u03ba\u03ae \u03ba\u03b1\u03c4\u03b5\u03cd\u03b8\u03c5\u03bd\u03c3\u03b7.<\/p>\n<p>\u0397 \u03bc\u03b5\u03c4\u03c1\u03b9\u03ba\u03ae \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c0\u03ac\u03bd\u03c4\u03b1 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03b2\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03bc\u03b1\u03b6\u03af \u03bc\u03b5 \u03c4\u03b7\u03bd \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1. \u0388\u03bd\u03b1 baseline \u03ba\u03bf\u03bd\u03c4\u03ac \u03c3\u03c4\u03bf \u03c4\u03b1\u03b2\u03ac\u03bd\u03b9 \u03b1\u03c6\u03ae\u03bd\u03b5\u03b9 \u03b5\u03bb\u03ac\u03c7\u03b9\u03c3\u03c4\u03bf \u03c0\u03b5\u03c1\u03b9\u03b8\u03ce\u03c1\u03b9\u03bf \u03b2\u03b5\u03bb\u03c4\u03af\u03c9\u03c3\u03b7\u03c2 \u03ba\u03b1\u03b9 \u03c3\u03c5\u03bc\u03c0\u03b9\u03ad\u03b6\u03b5\u03b9 \u03c4\u03bf TES. \u0391\u03bd\u03c4\u03af\u03b8\u03b5\u03c4\u03b1, \u03ad\u03bd\u03b1 baseline \u03ba\u03bf\u03bd\u03c4\u03ac \u03c3\u03c4\u03bf \u03bc\u03b7\u03b4\u03ad\u03bd \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c0\u03b1\u03c1\u03ac\u03b3\u03b5\u03b9 \u03b5\u03bb\u03ba\u03c5\u03c3\u03c4\u03b9\u03ba\u03cc \u03c3\u03c7\u03b5\u03c4\u03b9\u03ba\u03cc \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1, \u03b5\u03bd\u03ce \u03b7 \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03b5\u03b9 \u03b1\u03bd\u03b5\u03c0\u03b1\u03c1\u03ba\u03ae\u03c2 \u03b3\u03b9\u03b1 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03c7\u03c1\u03ae\u03c3\u03b7. \u03a4\u03bf TES \u03b2\u03bf\u03b7\u03b8\u03ac \u03c4\u03b7\u03bd \u03b1\u03c0\u03cc\u03c6\u03b1\u03c3\u03b7, \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03bd \u03b1\u03bd\u03c4\u03b9\u03ba\u03b1\u03b8\u03b9\u03c3\u03c4\u03ac \u03c4\u03bf business threshold \u03c4\u03b7\u03c2 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae\u03c2.<\/p>\n<h2 id=\"tes-delta-tes-a\">\u0394\u03cd\u03bf \u03c4\u03c1\u03cc\u03c0\u03bf\u03b9 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7\u03c2: TES-\u0394 \u03ba\u03b1\u03b9 TES-A<\/h2>\n<p>\u038c\u03c4\u03b1\u03bd \u03b7 \u03af\u03b4\u03b9\u03b1 \u03bf\u03b9\u03ba\u03bf\u03b3\u03ad\u03bd\u03b5\u03b9\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03c9\u03bd \u03c0\u03c1\u03bf\u03c3\u03c6\u03ad\u03c1\u03b5\u03b9 reasoning toggle, \u03bf\u03b9 \u03b5\u03c1\u03b5\u03c5\u03bd\u03b7\u03c4\u03ad\u03c2 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03bf\u03cd\u03bd \u03c4\u03bf TES-\u0394. \u0397 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7 \u03b3\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03c4\u03b7 reasoning \u03ba\u03b1\u03b9 \u03c4\u03b7 non-reasoning \u03b5\u03ba\u03b4\u03bf\u03c7\u03ae \u03c4\u03b7\u03c2 \u03af\u03b4\u03b9\u03b1\u03c2 \u03bf\u03b9\u03ba\u03bf\u03b3\u03ad\u03bd\u03b5\u03b9\u03b1\u03c2, \u03ac\u03c1\u03b1 \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b1 \u03b1\u03c1\u03c7\u03b9\u03c4\u03b5\u03ba\u03c4\u03bf\u03bd\u03b9\u03ba\u03ad\u03c2 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ad\u03c2. \u0393\u03b9\u03b1 \u03ba\u03bb\u03b5\u03b9\u03c3\u03c4\u03ad\u03c2 \u03bf\u03b9\u03ba\u03bf\u03b3\u03ad\u03bd\u03b5\u03b9\u03b5\u03c2, \u03b1\u03ba\u03cc\u03bc\u03b7 \u03ba\u03b1\u03b9 \u03b1\u03c5\u03c4\u03ae \u03b7 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03b2\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c9\u03c2 vendor-paired \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03c9\u03c2 \u03b1\u03c0\u03cc\u03b4\u03b5\u03b9\u03be\u03b7 \u03cc\u03c4\u03b9 \u03ac\u03bb\u03bb\u03b1\u03be\u03b5 \u03bc\u03cc\u03bd\u03bf \u03ad\u03bd\u03b1\u03c2 \u03c0\u03b1\u03c1\u03ac\u03b3\u03bf\u03bd\u03c4\u03b1\u03c2.<\/p>\n<p>\u0393\u03b9\u03b1 frontier \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03c7\u03c9\u03c1\u03af\u03c2 \u03ac\u03bc\u03b5\u03c3\u03bf non-reasoning \u03b4\u03af\u03b4\u03c5\u03bc\u03bf \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af\u03c4\u03b1\u03b9 \u03c4\u03bf TES-A. \u03a4\u03bf baseline \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf \u03b4\u03b9\u03b1\u03b8\u03ad\u03c3\u03b9\u03bc\u03bf non-reasoning \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c3\u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf benchmark. \u0391\u03c5\u03c4\u03cc \u03b1\u03bd\u03c4\u03b1\u03bd\u03b1\u03ba\u03bb\u03ac \u03bc\u03b9\u03b1 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b1\u03b3\u03bf\u03c1\u03b1\u03c3\u03c4\u03b9\u03ba\u03ae \u03b1\u03c0\u03cc\u03c6\u03b1\u03c3\u03b7, \u03b1\u03bb\u03bb\u03ac \u03c3\u03c5\u03b3\u03c7\u03ad\u03b5\u03b9 \u03c4\u03bf reasoning \u03bc\u03b5 \u03b3\u03b5\u03bd\u03b9\u03ba\u03cc\u03c4\u03b5\u03c1\u03b5\u03c2 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ad\u03c2 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03b5 \u03bf\u03b9\u03ba\u03bf\u03b3\u03ad\u03bd\u03b5\u03b9\u03b5\u03c2. \u03a3\u03c4\u03b7 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7, \u03c4\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 Gemini \u03b1\u03bd\u03ae\u03ba\u03bf\u03c5\u03bd \u03c3\u03b5 \u03b1\u03c5\u03c4\u03ae \u03c4\u03b7\u03bd \u03ba\u03b1\u03c4\u03b7\u03b3\u03bf\u03c1\u03af\u03b1 \u03ba\u03b1\u03b9 \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03bd\u03c4\u03b9\u03bc\u03b5\u03c4\u03c9\u03c0\u03af\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c3\u03b1\u03bd \u03b5\u03bb\u03b5\u03b3\u03c7\u03cc\u03bc\u03b5\u03bd\u03bf \u03c0\u03b5\u03af\u03c1\u03b1\u03bc\u03b1 toggle.<\/p>\n<h2 id=\"domi-ergasias\">\u0397 \u03b4\u03bf\u03bc\u03ae \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf \u03b1\u03c0\u03cc \u03c4\u03b7 \u03b4\u03c5\u03c3\u03ba\u03bf\u03bb\u03af\u03b1<\/h2>\n<p>\u039f\u03b9 \u03b5\u03c0\u03c4\u03ac \u03b4\u03bf\u03ba\u03b9\u03bc\u03ad\u03c2 \u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03bf\u03c5\u03bd instruction following, knowledge recall, \u03b4\u03b9\u03b1\u03b4\u03bf\u03c7\u03b9\u03ba\u03ae \u03bb\u03bf\u03b3\u03b9\u03ba\u03ae \u03b5\u03c0\u03b5\u03be\u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03ba\u03b1\u03b9 frontier-domain reasoning. \u03a4\u03b1 AIME 2025 \u03ba\u03b1\u03b9 LiveCodeBench, \u03cc\u03c0\u03bf\u03c5 \u03ba\u03ac\u03b8\u03b5 \u03b5\u03bd\u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03bf \u03b2\u03ae\u03bc\u03b1 \u03b5\u03c0\u03b7\u03c1\u03b5\u03ac\u03b6\u03b5\u03b9 \u03c4\u03bf \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03bf, \u03b4\u03b5\u03af\u03c7\u03bd\u03bf\u03c5\u03bd \u03b9\u03c3\u03c7\u03c5\u03c1\u03ae \u03b1\u03c0\u03cc\u03b4\u03bf\u03c3\u03b7 \u03b1\u03c0\u03cc \u03c4\u03bf reasoning. \u03a4\u03bf \u03af\u03b4\u03b9\u03bf \u03b9\u03c3\u03c7\u03cd\u03b5\u03b9 \u03ba\u03b1\u03c4\u03ac \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf \u03b3\u03b9\u03b1 \u03c4\u03bf IFBench. \u03a0\u03c1\u03cc\u03ba\u03b5\u03b9\u03c4\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b5\u03c2 \u03cc\u03c0\u03c9\u03c2 \u03b1\u03c0\u03cc\u03b4\u03b5\u03b9\u03be\u03b7, \u03c3\u03cd\u03bd\u03b8\u03b5\u03c3\u03b7 \u03ba\u03b1\u03b9 debugging \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1 \u03ae \u03b5\u03ba\u03c4\u03ad\u03bb\u03b5\u03c3\u03b7 \u03c0\u03bf\u03bb\u03bb\u03ce\u03bd \u03b1\u03bb\u03bb\u03b7\u03bb\u03b5\u03be\u03b1\u03c1\u03c4\u03ce\u03bc\u03b5\u03bd\u03c9\u03bd \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03ce\u03bd.<\/p>\n<p>\u03a3\u03c4\u03bf\u03bd \u03b1\u03bd\u03c4\u03af\u03c0\u03bf\u03b4\u03b1, \u03c4\u03bf MMLU-Pro \u03ba\u03b1\u03b9 \u03c4\u03bf GPQA Diamond \u03b5\u03bc\u03c6\u03b1\u03bd\u03af\u03b6\u03bf\u03c5\u03bd \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf TES, \u03c0\u03b1\u03c1\u03cc\u03c4\u03b9 \u03b8\u03b5\u03c9\u03c1\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03b4\u03cd\u03c3\u03ba\u03bf\u03bb\u03b1 benchmarks. \u0397 \u03b1\u03bd\u03ac\u03ba\u03bb\u03b7\u03c3\u03b7 \u03b3\u03bd\u03ce\u03c3\u03b7\u03c2 \u03b4\u03b5\u03bd \u03b2\u03b5\u03bb\u03c4\u03b9\u03ce\u03bd\u03b5\u03c4\u03b1\u03b9 \u03b1\u03bd\u03b1\u03b3\u03ba\u03b1\u03c3\u03c4\u03b9\u03ba\u03ac \u03bc\u03b5 \u03bc\u03b5\u03b3\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b7 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03ae \u03b4\u03b9\u03b1\u03b4\u03c1\u03bf\u03bc\u03ae: \u03b1\u03bd \u03bb\u03b5\u03af\u03c0\u03b5\u03b9 \u03c4\u03bf \u03c3\u03c9\u03c3\u03c4\u03cc \u03b3\u03b5\u03b3\u03bf\u03bd\u03cc\u03c2, \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b1 reasoning tokens \u03b4\u03b5\u03bd \u03c4\u03bf \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03bf\u03cd\u03bd \u03b1\u03be\u03b9\u03cc\u03c0\u03b9\u03c3\u03c4\u03b1. \u03a4\u03b1 HLE \u03ba\u03b1\u03b9 CritPt \u03b2\u03c1\u03af\u03c3\u03ba\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c3\u03b5 frontier \u03c0\u03b5\u03c1\u03b9\u03bf\u03c7\u03ad\u03c2, \u03cc\u03c0\u03bf\u03c5 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c0\u03b1\u03c1\u03ac\u03bb\u03bb\u03b7\u03bb\u03b7 \u03b5\u03be\u03ad\u03c4\u03b1\u03c3\u03b7 TES \u03ba\u03b1\u03b9 \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03b7\u03c2 \u03b5\u03c0\u03af\u03b4\u03bf\u03c3\u03b7\u03c2, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03c4\u03b1 baselines \u03ad\u03c7\u03bf\u03c5\u03bd \u03c0\u03bf\u03bb\u03cd \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03cc \u03b4\u03b9\u03b1\u03b8\u03ad\u03c3\u03b9\u03bc\u03bf \u03c0\u03b5\u03c1\u03b9\u03b8\u03ce\u03c1\u03b9\u03bf.<\/p>\n<p>\u0393\u03b9\u03b1 \u03bc\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03af\u03c1\u03b7\u03c3\u03b7 \u03b1\u03c5\u03c4\u03cc \u03bc\u03b5\u03c4\u03b1\u03c6\u03c1\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03c3\u03b1\u03c6\u03ae \u03b4\u03c1\u03bf\u03bc\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7. Planning, \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03bf\u03af, \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1 \u03ba\u03b1\u03b9 verification \u03b5\u03af\u03bd\u03b1\u03b9 \u03bb\u03bf\u03b3\u03b9\u03ba\u03bf\u03af \u03c5\u03c0\u03bf\u03c8\u03ae\u03c6\u03b9\u03bf\u03b9 \u03b3\u03b9\u03b1 reasoning. \u0391\u03bd\u03c4\u03af\u03b8\u03b5\u03c4\u03b1, retrieval, classification, routing, \u03bc\u03bf\u03c1\u03c6\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03ae \u03c3\u03cd\u03bd\u03c4\u03bf\u03bc\u03b7 \u03c0\u03b5\u03c1\u03af\u03bb\u03b7\u03c8\u03b7 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c0\u03c1\u03ce\u03c4\u03b1 \u03b4\u03bf\u03ba\u03b9\u03bc\u03ae \u03bc\u03b5 \u03ad\u03bd\u03b1 \u03b9\u03c3\u03c7\u03c5\u03c1\u03cc non-reasoning \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf. \u0397 \u03b4\u03c5\u03c3\u03ba\u03bf\u03bb\u03af\u03b1 \u03c0\u03bf\u03c5 \u03b1\u03b9\u03c3\u03b8\u03ac\u03bd\u03b5\u03c4\u03b1\u03b9 \u03bf \u03ac\u03bd\u03b8\u03c1\u03c9\u03c0\u03bf\u03c2 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03c0\u03cc \u03bc\u03cc\u03bd\u03b7 \u03c4\u03b7\u03c2 \u03b5\u03c0\u03b1\u03c1\u03ba\u03ad\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf.<\/p>\n<div class=\"td-comparison-cards\">\n<article class=\"td-platform-card\">\n<h3>Sequential inference<\/h3>\n<p>\u039c\u03b1\u03b8\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b1\u03c0\u03bf\u03b4\u03b5\u03af\u03be\u03b5\u03b9\u03c2, \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1 \u03ba\u03b1\u03b9 debugging \u03ba\u03b5\u03c1\u03b4\u03af\u03b6\u03bf\u03c5\u03bd \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf \u03cc\u03c4\u03b1\u03bd \u03ba\u03ac\u03b8\u03b5 \u03b5\u03bd\u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03bf \u03b2\u03ae\u03bc\u03b1 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03af\u03b6\u03b5\u03b9 \u03c4\u03bf \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03bf.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">AIME 2025<\/span><span class=\"td-badge\">LiveCodeBench<\/span><\/div>\n<\/article>\n<article class=\"td-platform-card td-platform-card--navy\">\n<h3>Knowledge recall<\/h3>\n<p>\u0397 \u03bc\u03b5\u03b3\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b7 \u03b1\u03bb\u03c5\u03c3\u03af\u03b4\u03b1 \u03c3\u03ba\u03ad\u03c8\u03b7\u03c2 \u03b4\u03b5\u03bd \u03b1\u03bd\u03b1\u03c0\u03bb\u03b7\u03c1\u03ce\u03bd\u03b5\u03b9 \u03b1\u03be\u03b9\u03cc\u03c0\u03b9\u03c3\u03c4\u03b1 \u03ad\u03bd\u03b1 \u03b3\u03b5\u03b3\u03bf\u03bd\u03cc\u03c2 \u03c0\u03bf\u03c5 \u03bb\u03b5\u03af\u03c0\u03b5\u03b9 \u03b1\u03c0\u03cc \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03ba\u03b1\u03b9 \u03c3\u03c5\u03c7\u03bd\u03ac \u03b4\u03af\u03bd\u03b5\u03b9 \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf TES.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">MMLU-Pro<\/span><span class=\"td-badge\">GPQA Diamond<\/span><\/div>\n<\/article>\n<article class=\"td-platform-card\">\n<h3>Frontier difficulty<\/h3>\n<p>\u03a3\u03b5 HLE \u03ba\u03b1\u03b9 CritPt \u03c4\u03bf TES \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03b2\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03bc\u03b1\u03b6\u03af \u03bc\u03b5 \u03c4\u03b7\u03bd \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03c4\u03b1 baselines \u03ad\u03c7\u03bf\u03c5\u03bd \u03c0\u03bf\u03bb\u03cd \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03cc headroom.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">HLE<\/span><span class=\"td-badge\">CritPt<\/span><\/div>\n<\/article>\n<\/div>\n<h2 id=\"maximum-reasoning\">\u03a4\u03bf maximum reasoning \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03bf\u03c5\u03bc\u03c0\u03af \u00ab\u03bc\u03ad\u03b3\u03b9\u03c3\u03c4\u03b7\u03c2 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2\u00bb<\/h2>\n<p>\u03a3\u03b5 \u03ba\u03ac\u03b8\u03b5 \u03bf\u03b9\u03ba\u03bf\u03b3\u03ad\u03bd\u03b5\u03b9\u03b1 \u03c4\u03b7\u03c2 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7\u03c2 \u03c0\u03bf\u03c5 \u03c0\u03c1\u03bf\u03c3\u03c6\u03ad\u03c1\u03b5\u03b9 \u03c0\u03bf\u03bb\u03bb\u03b1\u03c0\u03bb\u03ac effort levels, \u03b7 \u03b1\u03c0\u03cc\u03b4\u03bf\u03c3\u03b7 \u03c6\u03b8\u03af\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b1\u03bd \u03c4\u03bf reasoning budget \u03b1\u03bd\u03b5\u03b2\u03b1\u03af\u03bd\u03b5\u03b9 \u03c0\u03ad\u03c1\u03b1 \u03b1\u03c0\u03cc \u03ad\u03bd\u03b1 \u03bc\u03ad\u03c4\u03c1\u03b9\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf. \u03a4\u03b1 \u03b5\u03c0\u03b9\u03c0\u03bb\u03ad\u03bf\u03bd tokens \u03b1\u03c5\u03be\u03ac\u03bd\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b3\u03c1\u03b7\u03b3\u03bf\u03c1\u03cc\u03c4\u03b5\u03c1\u03b1 \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1. \u03a3\u03b5 \u03bf\u03c1\u03b9\u03c3\u03bc\u03ad\u03bd\u03b5\u03c2 \u03bc\u03b5\u03c4\u03b1\u03b2\u03ac\u03c3\u03b5\u03b9\u03c2 \u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03c0\u03ad\u03c6\u03c4\u03b5\u03b9: \u03c4\u03bf GPT-5.2 \u03c3\u03c4\u03bf LiveCodeBench \u03bc\u03b5\u03b9\u03ce\u03b8\u03b7\u03ba\u03b5 \u03ba\u03b1\u03c4\u03ac 0,5 \u03c0\u03bf\u03c3\u03bf\u03c3\u03c4\u03b9\u03b1\u03af\u03b1 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b1 \u03b1\u03c0\u03cc medium \u03c3\u03b5 xhigh, \u03b5\u03bd\u03ce \u03c4\u03bf DeepSeek V4 Pro \u03c3\u03c4\u03bf GPQA Diamond \u03bc\u03b5\u03b9\u03ce\u03b8\u03b7\u03ba\u03b5 \u03ba\u03b1\u03c4\u03ac 1,7 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b5\u03c2 \u03b1\u03c0\u03cc high \u03c3\u03b5 max.<\/p>\n<p>\u0386\u03bb\u03bb\u03b5\u03c2 \u03bc\u03b5\u03c4\u03b1\u03b2\u03ac\u03c3\u03b5\u03b9\u03c2 \u03ad\u03c7\u03bf\u03c5\u03bd \u03b8\u03b5\u03c4\u03b9\u03ba\u03cc \u03b1\u03bb\u03bb\u03ac \u03bc\u03b9\u03ba\u03c1\u03cc \u03ba\u03ad\u03c1\u03b4\u03bf\u03c2. \u03a3\u03c4\u03bf GPQA Diamond, \u03c4\u03bf GPT-5.5 \u03ba\u03ad\u03c1\u03b4\u03b9\u03c3\u03b5 0,3 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b1 \u03b1\u03c0\u03cc high \u03c3\u03b5 xhigh, \u03bc\u03b5 marginal TES 0,175. \u03a3\u03c4\u03bf AIME 2025, \u03c4\u03bf GPT-5.2 \u03ba\u03ad\u03c1\u03b4\u03b9\u03c3\u03b5 2,3 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b5\u03c2 \u03ba\u03b1\u03b9 \u03b4\u03b9\u03b1\u03c4\u03ae\u03c1\u03b7\u03c3\u03b5 marginal TES 1,022, \u03b4\u03b5\u03af\u03c7\u03bd\u03bf\u03bd\u03c4\u03b1\u03c2 \u03cc\u03c4\u03b9 \u03b7 \u03b4\u03b9\u03b1\u03b4\u03bf\u03c7\u03b9\u03ba\u03ae \u03bc\u03b1\u03b8\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b5\u03c0\u03b5\u03be\u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03b1\u03bd\u03c4\u03ad\u03c7\u03b5\u03b9 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b1 \u03c4\u03bf \u03c5\u03c8\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf effort. \u03a4\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u00ab\u03c0\u03ac\u03bd\u03c4\u03b1 \u03bb\u03b9\u03b3\u03cc\u03c4\u03b5\u03c1\u03bf reasoning\u00bb, \u03b1\u03bb\u03bb\u03ac \u00ab\u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf \u03bc\u03cc\u03bd\u03bf \u03cc\u03c0\u03bf\u03c5 \u03b1\u03c0\u03bf\u03b4\u03b5\u03b4\u03b5\u03b9\u03b3\u03bc\u03ad\u03bd\u03b1 \u03c0\u03bb\u03b7\u03c1\u03ce\u03bd\u03b5\u03b9\u00bb.<\/p>\n<p>\u0397 \u03b1\u03c3\u03c6\u03b1\u03bb\u03ae\u03c2 \u03b4\u03b9\u03b1\u03b4\u03b9\u03ba\u03b1\u03c3\u03af\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03bd\u03b1 \u03be\u03b5\u03ba\u03b9\u03bd\u03ac \u03ad\u03bd\u03b1 team \u03bc\u03b5 medium \u03ae high effort \u03c3\u03b5 \u03b1\u03bd\u03c4\u03b9\u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03b5\u03c5\u03c4\u03b9\u03ba\u03cc validation set, \u03bd\u03b1 \u03c3\u03c5\u03b3\u03ba\u03c1\u03af\u03bd\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b7 \u03b2\u03b1\u03b8\u03bc\u03af\u03b4\u03b1 \u03ba\u03b1\u03b9 \u03bd\u03b1 \u03b1\u03bd\u03b5\u03b2\u03b1\u03af\u03bd\u03b5\u03b9 \u03bc\u03cc\u03bd\u03bf \u03b1\u03bd \u03b7 \u03c0\u03c1\u03cc\u03c3\u03b8\u03b5\u03c4\u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03b5\u03b9 tokens \u03ba\u03b1\u03b9 latency. \u03a4\u03bf maximum effort \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b5\u03b9\u03b4\u03b9\u03ba\u03ae \u03c4\u03b5\u03ba\u03bc\u03b7\u03c1\u03af\u03c9\u03c3\u03b7 \u03b3\u03b9\u03b1 \u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf task \u03ba\u03b1\u03b9 model, \u03cc\u03c7\u03b9 \u03b1\u03c0\u03bb\u03ae \u03c5\u03c0\u03cc\u03b8\u03b5\u03c3\u03b7 \u03cc\u03c4\u03b9 \u03ad\u03bd\u03b1 \u03b4\u03cd\u03c3\u03ba\u03bf\u03bb\u03bf \u03b1\u03af\u03c4\u03b7\u03bc\u03b1 \u03b1\u03c0\u03b1\u03b9\u03c4\u03b5\u03af \u03c4\u03b7 \u03bc\u03b5\u03b3\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b7 \u03b4\u03b9\u03b1\u03b8\u03ad\u03c3\u03b9\u03bc\u03b7 \u03c3\u03ba\u03ad\u03c8\u03b7.<\/p>\n<div class=\"td-decision-band\">\n<p class=\"td-decision-kicker\">\u039a\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03ba\u03bb\u03b9\u03bc\u03ac\u03ba\u03c9\u03c3\u03b7\u03c2<\/p>\n<p class=\"td-decision-title\"><strong>\u039c\u03b7\u03bd \u03b1\u03b3\u03bf\u03c1\u03ac\u03b6\u03b5\u03c4\u03b5 maximum effort \u03c7\u03c9\u03c1\u03af\u03c2 marginal test<\/strong><\/p>\n<p>\u0391\u03bd\u03b5\u03b2\u03ac\u03c3\u03c4\u03b5 \u03c4\u03bf reasoning effort \u03bc\u03cc\u03bd\u03bf \u03cc\u03c4\u03b1\u03bd \u03b7 \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b7 \u03b2\u03b1\u03b8\u03bc\u03af\u03b4\u03b1 \u03b2\u03b5\u03bb\u03c4\u03b9\u03ce\u03bd\u03b5\u03b9 \u03c4\u03bf \u03b4\u03b9\u03ba\u03cc \u03c3\u03b1\u03c2 quality metric \u03b1\u03c1\u03ba\u03b5\u03c4\u03ac \u03ce\u03c3\u03c4\u03b5 \u03bd\u03b1 \u03b4\u03b9\u03ba\u03b1\u03b9\u03bf\u03bb\u03bf\u03b3\u03b5\u03af \u03c4\u03b1 \u03c0\u03c1\u03cc\u03c3\u03b8\u03b5\u03c4\u03b1 generated tokens, \u03c4\u03bf latency \u03ba\u03b1\u03b9 \u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf\u03c5 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5.<\/p>\n<\/div>\n<h2 id=\"reasoning-cost-share\">Reasoning Cost Share: \u03c0\u03bf\u03cd \u03c0\u03b7\u03b3\u03b1\u03af\u03bd\u03b5\u03b9 \u03bf \u03bb\u03bf\u03b3\u03b1\u03c1\u03b9\u03b1\u03c3\u03bc\u03cc\u03c2<\/h2>\n<p>\u03a4\u03bf Reasoning Cost Share, \u03ae RCS, \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03b5\u03b9 \u03c4\u03bf \u03c0\u03bf\u03c3\u03bf\u03c3\u03c4\u03cc \u03c4\u03bf\u03c5 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03bf\u03cd inference cost \u03c0\u03bf\u03c5 \u03ba\u03b1\u03c4\u03b1\u03bd\u03b1\u03bb\u03ce\u03bd\u03b5\u03b9 \u03b7 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03ae \u03c3\u03ba\u03ad\u03c8\u03b7. \u0394\u03b5\u03bd \u03bc\u03c0\u03b1\u03af\u03bd\u03b5\u03b9 \u03c3\u03c4\u03bf\u03bd \u03c4\u03cd\u03c0\u03bf \u03c4\u03bf\u03c5 TES, \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03c4\u03b7 \u03c3\u03cd\u03bd\u03b8\u03b5\u03c3\u03b7 \u03c4\u03bf\u03c5 \u03ba\u03cc\u03c3\u03c4\u03bf\u03c5\u03c2. \u03a3\u03c4\u03bf \u03c3\u03cd\u03bd\u03bf\u03bb\u03bf \u03c4\u03c9\u03bd reasoning runs \u03c4\u03b7\u03c2 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7\u03c2, \u03b7 \u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03b7 \u03c4\u03b9\u03bc\u03ae RCS \u03ae\u03c4\u03b1\u03bd 94,7%, \u03b5\u03bd\u03ce \u03b1\u03c1\u03ba\u03b5\u03c4\u03ad\u03c2 frontier \u03b5\u03ba\u03c4\u03b5\u03bb\u03ad\u03c3\u03b5\u03b9\u03c2 \u03be\u03b5\u03c0\u03ad\u03c1\u03b1\u03c3\u03b1\u03bd \u03c4\u03bf 99%. \u039c\u03b5 \u03ac\u03bb\u03bb\u03b1 \u03bb\u03cc\u03b3\u03b9\u03b1, \u03c4\u03bf \u03bc\u03b5\u03b3\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf \u03bc\u03ad\u03c1\u03bf\u03c2 \u03c4\u03b7\u03c2 \u03b4\u03b1\u03c0\u03ac\u03bd\u03b7\u03c2 \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03c3\u03b5 \u03bd\u03b1 \u03b1\u03c6\u03bf\u03c1\u03ac \u03c4\u03b7 \u03b4\u03b9\u03b1\u03b4\u03b9\u03ba\u03b1\u03c3\u03af\u03b1 \u03c0\u03c1\u03b9\u03bd \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03bf\u03c1\u03b1\u03c4\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7.<\/p>\n<div class=\"td-chart td-chart--metrics\" role=\"img\" aria-label=\"\u03a3\u03cd\u03bd\u03bf\u03c8\u03b7 \u03c4\u03b7\u03c2 \u03ad\u03c1\u03b5\u03c5\u03bd\u03b1\u03c2: 151 model-benchmark runs, 7 benchmarks, 77 reasoning \u03ba\u03b1\u03b9 non-reasoning \u03c3\u03c5\u03b3\u03ba\u03c1\u03af\u03c3\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 \u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03bf Reasoning Cost Share 94,7 \u03c4\u03bf\u03b9\u03c2 \u03b5\u03ba\u03b1\u03c4\u03cc\">\n<div class=\"td-chart-head\">\n<p class=\"td-chart-title\">\u03a4\u03bf \u03b1\u03c0\u03bf\u03c4\u03cd\u03c0\u03c9\u03bc\u03b1 \u03c4\u03b7\u03c2 \u03ad\u03c1\u03b5\u03c5\u03bd\u03b1\u03c2<\/p>\n<p class=\"td-chart-subtitle\">\u039f\u03b9 \u03b1\u03c1\u03b9\u03b8\u03bc\u03bf\u03af \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03bf\u03c5\u03bd \u03c4\u03bf dataset \u03ba\u03b1\u03b9 \u03c4\u03b7 \u03c3\u03cd\u03bd\u03b8\u03b5\u03c3\u03b7 \u03ba\u03cc\u03c3\u03c4\u03bf\u03c5\u03c2 \u03c4\u03b7\u03c2 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7\u03c2, \u03cc\u03c7\u03b9 \u03b5\u03b3\u03b3\u03cd\u03b7\u03c3\u03b7 \u03b1\u03c0\u03cc\u03b4\u03bf\u03c3\u03b7\u03c2 \u03b3\u03b9\u03b1 \u03ba\u03ac\u03b8\u03b5 \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03b9\u03ba\u03cc workload.<\/p>\n<\/div>\n<div class=\"td-chart-body\">\n<div class=\"td-metric-grid td-metric-grid--four-columns\">\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">151<\/span><span class=\"td-metric-label\">model-benchmark runs<\/span><span class=\"td-metric-note\">27 \u03b4\u03b9\u03b1\u03bc\u03bf\u03c1\u03c6\u03ce\u03c3\u03b5\u03b9\u03c2 \u03b1\u03c0\u03cc 8 \u03bf\u03b9\u03ba\u03bf\u03b3\u03ad\u03bd\u03b5\u03b9\u03b5\u03c2<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">7<\/span><span class=\"td-metric-label\">benchmarks<\/span><span class=\"td-metric-note\">\u03a4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b9\u03c2 \u03b4\u03bf\u03bc\u03ad\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">77<\/span><span class=\"td-metric-label\">TES \u03c3\u03c5\u03b3\u03ba\u03c1\u03af\u03c3\u03b5\u03b9\u03c2<\/span><span class=\"td-metric-note\">Reasoning \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 non-reasoning baseline<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">94,7%<\/span><span class=\"td-metric-label\">\u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03bf RCS<\/span><span class=\"td-metric-note\">\u039c\u03b5\u03c1\u03af\u03b4\u03b9\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c5\u03c2 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03ae\u03c2 \u03c3\u03ba\u03ad\u03c8\u03b7\u03c2<\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<p>\u03a4\u03bf RCS \u03b1\u03c5\u03be\u03ac\u03bd\u03b5\u03c4\u03b1\u03b9 \u03c3\u03c5\u03bd\u03ae\u03b8\u03c9\u03c2 \u03bc\u03b1\u03b6\u03af \u03bc\u03b5 \u03c4\u03bf effort. \u0393\u03b9\u03b1 \u03c4\u03bf GPT-5.5 \u03c3\u03c4\u03bf GPQA Diamond, \u03b1\u03bd\u03ad\u03b2\u03b7\u03ba\u03b5 \u03b1\u03c0\u03cc 76,7% \u03c3\u03c4\u03bf high \u03c3\u03b5 85,5% \u03c3\u03c4\u03bf xhigh. \u0391\u03c5\u03c4\u03cc \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03bf \u03b3\u03b9\u03b1 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ad\u03c2 \u03c0\u03bf\u03c5 \u03b5\u03c0\u03b9\u03c3\u03c4\u03c1\u03ad\u03c6\u03bf\u03c5\u03bd \u03bc\u03b9\u03ba\u03c1\u03cc \u03c4\u03b5\u03bb\u03b9\u03ba\u03cc \u03ba\u03b5\u03af\u03bc\u03b5\u03bd\u03bf: \u03bc\u03b9\u03b1 \u03bb\u03b9\u03c4\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u03b4\u03b5\u03bd \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03b7 \u03ba\u03bb\u03ae\u03c3\u03b7 \u03ae\u03c4\u03b1\u03bd \u03c6\u03b8\u03b7\u03bd\u03ae. \u0397 \u03c0\u03b1\u03c1\u03b1\u03ba\u03bf\u03bb\u03bf\u03cd\u03b8\u03b7\u03c3\u03b7 \u03bc\u03cc\u03bd\u03bf \u03c4\u03c9\u03bd output tokens \u03c0\u03bf\u03c5 \u03b2\u03bb\u03ad\u03c0\u03b5\u03b9 \u03bf \u03c7\u03c1\u03ae\u03c3\u03c4\u03b7\u03c2 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c5\u03c0\u03bf\u03c4\u03b9\u03bc\u03ac \u03b4\u03c1\u03b1\u03c3\u03c4\u03b9\u03ba\u03ac \u03c4\u03bf \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2.<\/p>\n<h2 id=\"cloud-on-premises\">Cloud \u03ae on-premises: \u03c4\u03bf \u03af\u03b4\u03b9\u03bf TES, \u03ac\u03bb\u03bb\u03b7 \u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ae \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc\u03c4\u03b7\u03c4\u03b1<\/h2>\n<p>\u03a4\u03bf TES \u03b2\u03b1\u03c3\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03bb\u03cc\u03b3\u03bf\u03c5\u03c2 tokens \u03ba\u03b1\u03b9 \u03b4\u03b5\u03bd \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03bc\u03b5 \u03c4\u03b7\u03bd \u03c4\u03b9\u03bc\u03ae \u03c4\u03bf\u03c5 \u03c0\u03b1\u03c1\u03cc\u03c7\u03bf\u03c5. \u03a4\u03bf \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2, \u03cc\u03bc\u03c9\u03c2, \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03ad\u03bd\u03c4\u03bf\u03bd\u03b1 \u03b1\u03bd\u03ac\u03bb\u03bf\u03b3\u03b1 \u03bc\u03b5 \u03c4\u03bf deployment. \u039f\u03b9 \u03b5\u03c1\u03b5\u03c5\u03bd\u03b7\u03c4\u03ad\u03c2 \u03c5\u03c0\u03bf\u03bb\u03cc\u03b3\u03b9\u03c3\u03b1\u03bd \u03c4\u03bf Deployment Cost Multiplier, \u03c3\u03c5\u03b3\u03ba\u03c1\u03af\u03bd\u03bf\u03bd\u03c4\u03b1\u03c2 cloud API \u03bc\u03b5 \u03b4\u03b9\u03ba\u03ad\u03c2 \u03c4\u03bf\u03c5\u03c2 \u03b5\u03ba\u03c4\u03b5\u03bb\u03ad\u03c3\u03b5\u03b9\u03c2 \u03c3\u03b5 \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 8x NVIDIA B300, FP16. \u03a4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03bf\u03cd \u03ba\u03cc\u03c3\u03c4\u03bf\u03c5\u03c2 \u03b8\u03b5\u03c9\u03c1\u03b5\u03af \u03c0\u03b5\u03bd\u03c4\u03b1\u03b5\u03c4\u03ae \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03af\u03b1, \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 1,013 \u03b5\u03ba\u03b1\u03c4. \u03b4\u03bf\u03bb\u03ac\u03c1\u03b9\u03b1 \u03ba\u03b1\u03b9 \u03b1\u03c0\u03bf\u03c3\u03b2\u03b5\u03c3\u03bc\u03ad\u03bd\u03bf \u03c1\u03c5\u03b8\u03bc\u03cc 0,00633 \u03b4\u03bf\u03bb\u03ac\u03c1\u03b9\u03b1 \u03b1\u03bd\u03ac \u03b4\u03b5\u03c5\u03c4\u03b5\u03c1\u03cc\u03bb\u03b5\u03c0\u03c4\u03bf.<\/p>\n<p>\u03a3\u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf setup, \u03c4\u03bf Qwen3.5-397B-A17B \u03b5\u03bc\u03c6\u03ac\u03bd\u03b9\u03c3\u03b5 DCM 25,6x, \u03c4\u03bf Qwen3.6-27B 20,0x \u03ba\u03b1\u03b9 \u03c4\u03bf Qwen3-235B-A22B 13,6x. \u039c\u03b9\u03ba\u03c1\u03cc\u03c4\u03b5\u03c1\u03b5\u03c2 dense \u03b4\u03b9\u03b1\u03bc\u03bf\u03c1\u03c6\u03ce\u03c3\u03b5\u03b9\u03c2 \u03b5\u03af\u03c7\u03b1\u03bd \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf\u03c5\u03c2 \u03c0\u03bf\u03bb\u03bb\u03b1\u03c0\u03bb\u03b1\u03c3\u03b9\u03b1\u03c3\u03c4\u03ad\u03c2, \u03cc\u03c0\u03c9\u03c2 2,8x \u03b3\u03b9\u03b1 \u03c4\u03bf Qwen3-32B \u03ba\u03b1\u03b9 2,0x \u03b3\u03b9\u03b1 \u03c4\u03bf Gemma 4 31B. \u039f\u03b9 MoE \u03b1\u03c1\u03c7\u03b9\u03c4\u03b5\u03ba\u03c4\u03bf\u03bd\u03b9\u03ba\u03ad\u03c2 \u03b5\u03c0\u03c9\u03c6\u03b5\u03bb\u03ae\u03b8\u03b7\u03ba\u03b1\u03bd \u03b1\u03c0\u03cc sparse activation \u03ba\u03b1\u03b9 \u03c5\u03c8\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf throughput, \u03b5\u03bd\u03ce \u03c3\u03c4\u03b1 dense \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03b7 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03b5\u03be\u03b1\u03c1\u03c4\u03b9\u03cc\u03c4\u03b1\u03bd \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03b5\u03bc\u03c0\u03bf\u03c1\u03b9\u03ba\u03ae \u03c4\u03b9\u03bc\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7.<\/p>\n<p>\u0391\u03c5\u03c4\u03ac \u03c4\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03b9\u03c3\u03c7\u03cd\u03bf\u03c5\u03bd \u03c5\u03c0\u03cc \u03c5\u03c8\u03b7\u03bb\u03ae \u03b1\u03be\u03b9\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03c4\u03bf\u03c5 \u03b5\u03be\u03bf\u03c0\u03bb\u03b9\u03c3\u03bc\u03bf\u03cd. \u03a3\u03b5 \u03bc\u03b9\u03ba\u03c1\u03cc, \u03b1\u03c3\u03c4\u03b1\u03b8\u03ad\u03c2 \u03ae \u03b5\u03c0\u03bf\u03c7\u03b9\u03ba\u03cc \u03c6\u03cc\u03c1\u03c4\u03bf, \u03c4\u03bf \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc on-premises \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03b1\u03bd\u03ac token \u03b1\u03bd\u03b5\u03b2\u03b1\u03af\u03bd\u03b5\u03b9 \u03ba\u03b1\u03b9 \u03c4\u03bf cloud \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c0\u03b1\u03c1\u03b1\u03bc\u03b5\u03af\u03bd\u03b5\u03b9 \u03c0\u03c1\u03bf\u03c4\u03b9\u03bc\u03cc\u03c4\u03b5\u03c1\u03bf. \u03a4\u03bf DCM \u03b4\u03b5\u03bd \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03ad\u03bd\u03b1 \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc TES \u03c3\u03b5 \u03c5\u03c8\u03b7\u03bb\u03cc\u0387 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03cc\u03bc\u03c9\u03c2 \u03bd\u03b1 \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c8\u03b5\u03b9 \u03bc\u03b9\u03b1 \u03b1\u03ba\u03c1\u03b9\u03b2\u03ae, \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b7 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03c3\u03b5 \u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac \u03b2\u03b9\u03ce\u03c3\u03b9\u03bc\u03b7.<\/p>\n<div class=\"td-article-note\"><strong>\u03a3\u03b7\u03bc\u03b1\u03bd\u03c4\u03b9\u03ba\u03ae \u03b4\u03b9\u03ac\u03ba\u03c1\u03b9\u03c3\u03b7:<\/strong> \u03c4\u03bf DCM \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03c4\u03bf \u03c0\u03bf\u03c3\u03cc \u03c0\u03bf\u03c5 \u03c0\u03bb\u03b7\u03c1\u03ce\u03bd\u03b5\u03c4\u03b5 \u03b3\u03b9\u03b1 \u03c4\u03bf \u03af\u03b4\u03b9\u03bf workload, \u03cc\u03c7\u03b9 \u03c4\u03bf TES \u03c4\u03bf\u03c5 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5. \u03a7\u03b1\u03bc\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf on-premises \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03b4\u03b5\u03bd \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03b1\u03c0\u03cc \u03bc\u03cc\u03bd\u03bf \u03c4\u03bf\u03c5 \u03ad\u03bd\u03b1 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03bf reasoning gain \u03c3\u03b5 \u03ba\u03b1\u03bb\u03cc \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1.<\/div>\n<h2 id=\"tessera-tetartimoria\">\u03a4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 \u03c4\u03b5\u03c4\u03b1\u03c1\u03c4\u03b7\u03bc\u03cc\u03c1\u03b9\u03b1 \u03b3\u03b9\u03b1 \u03c4\u03b7\u03bd \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5<\/h2>\n<p>\u0397 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03c3\u03c5\u03bd\u03b4\u03c5\u03ac\u03b6\u03b5\u03b9 TES \u03ba\u03b1\u03b9 cloud cost \u03c3\u03b5 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b9\u03c2 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c7\u03ad\u03c2. \u03a9\u03c2 \u03cc\u03c1\u03b9\u03bf \u03b1\u03c0\u03bf\u03b4\u03bf\u03c4\u03b9\u03ba\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af TES \u03c0\u03ac\u03bd\u03c9 \u03b1\u03c0\u03cc 1 \u03ba\u03b1\u03b9 \u03c9\u03c2 \u03cc\u03c1\u03b9\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c5\u03c2 \u03c4\u03b7 \u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03b7 \u03c4\u03b9\u03bc\u03ae 19,69 \u03b4\u03bf\u03bb\u03b1\u03c1\u03af\u03c9\u03bd \u03b1\u03bd\u03ac benchmark run \u03c3\u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf dataset. \u03a5\u03c8\u03b7\u03bb\u03cc TES \u03ba\u03b1\u03b9 \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03b5\u03c5\u03bd\u03bf\u03bf\u03cd\u03bd reasoning by default. \u03a5\u03c8\u03b7\u03bb\u03cc TES \u03ba\u03b1\u03b9 \u03c5\u03c8\u03b7\u03bb\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03bf\u03b4\u03b7\u03b3\u03bf\u03cd\u03bd \u03c3\u03b5 \u03b5\u03c0\u03b9\u03bb\u03b5\u03ba\u03c4\u03b9\u03ba\u03ae \u03c7\u03c1\u03ae\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03b5\u03be\u03ad\u03c4\u03b1\u03c3\u03b7 on-premises \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae\u03c2 \u03b3\u03b9\u03b1 \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03bf \u03c6\u03cc\u03c1\u03c4\u03bf.<\/p>\n<p>\u03a7\u03b1\u03bc\u03b7\u03bb\u03cc TES \u03ba\u03b1\u03b9 \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03bf\u03c5\u03bd \u03cc\u03c4\u03b9 \u03b7 \u03b1\u03c0\u03cc\u03c6\u03b1\u03c3\u03b7 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03be\u03b1\u03c1\u03c4\u03ac\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03b1\u03be\u03af\u03b1 \u03c4\u03b7\u03c2 \u03bc\u03b9\u03ba\u03c1\u03ae\u03c2 \u03b2\u03b5\u03bb\u03c4\u03af\u03c9\u03c3\u03b7\u03c2 \u03b3\u03b9\u03b1 \u03c4\u03b7\u03bd \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae. \u03a7\u03b1\u03bc\u03b7\u03bb\u03cc TES \u03ba\u03b1\u03b9 \u03c5\u03c8\u03b7\u03bb\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b7 \u03c0\u03b9\u03bf \u03ba\u03b1\u03b8\u03b1\u03c1\u03ae \u03c0\u03b5\u03c1\u03af\u03c0\u03c4\u03c9\u03c3\u03b7 \u03b3\u03b9\u03b1 \u03b1\u03c0\u03b5\u03bd\u03b5\u03c1\u03b3\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 reasoning \u03ae \u03bc\u03b5\u03af\u03c9\u03c3\u03b7 effort. \u0397 \u03c7\u03b1\u03c1\u03c4\u03bf\u03b3\u03c1\u03ac\u03c6\u03b7\u03c3\u03b7 \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03b5\u03af \u03ad\u03c4\u03bf\u03b9\u03bc\u03bf \u03c4\u03b9\u03bc\u03bf\u03ba\u03b1\u03c4\u03ac\u03bb\u03bf\u03b3\u03bf \u03b3\u03b9\u03b1 \u03ba\u03ac\u03b8\u03b5 \u03b5\u03c4\u03b1\u03b9\u03c1\u03b5\u03af\u03b1, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03c4\u03b1 thresholds \u03ba\u03b1\u03b9 \u03bf\u03b9 \u03c4\u03b9\u03bc\u03ad\u03c2 \u03b1\u03bb\u03bb\u03ac\u03b6\u03bf\u03c5\u03bd. \u0391\u03c0\u03bf\u03c4\u03b5\u03bb\u03b5\u03af \u03cc\u03bc\u03c9\u03c2 \u03c7\u03c1\u03ae\u03c3\u03b9\u03bc\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b1\u03c0\u03cc\u03c6\u03b1\u03c3\u03b7\u03c2 \u03c0\u03bf\u03c5 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03b1\u03c7\u03b8\u03b5\u03af \u03bc\u03b5 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03ac \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1.<\/p>\n<h2 id=\"ai-agents-workflows\">\u03a4\u03b9 \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03c3\u03b5 AI agents \u03ba\u03b1\u03b9 \u03c0\u03bf\u03bb\u03c5\u03b2\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac workflows<\/h2>\n<p>\u039f\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03b5\u03b9\u03c2 TES \u03c4\u03b7\u03c2 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 single-turn. \u03a3\u03b5 \u03ad\u03bd\u03b1\u03bd <a href=\"https:\/\/twodots.gr\/ai-agents-tool-calling-proto-lathos-alysida\/\">AI agent \u03bc\u03b5 \u03b4\u03b9\u03b1\u03b4\u03bf\u03c7\u03b9\u03ba\u03cc tool calling<\/a>, \u03b7 \u03ad\u03be\u03bf\u03b4\u03bf\u03c2 \u03b5\u03bd\u03cc\u03c2 \u03b2\u03ae\u03bc\u03b1\u03c4\u03bf\u03c2 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b3\u03af\u03bd\u03b5\u03b9 input \u03c3\u03c4\u03bf \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03bf. \u0391\u03bd \u03b4\u03b9\u03b1\u03c4\u03b7\u03c1\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 scratchpads, reasoning traces \u03ae \u03bc\u03b5\u03b3\u03ac\u03bb\u03b1 \u03b5\u03bd\u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03b1 outputs, \u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03b4\u03b5\u03bd \u03c0\u03c1\u03bf\u03c3\u03c4\u03af\u03b8\u03b5\u03c4\u03b1\u03b9 \u03b1\u03c0\u03bb\u03ce\u03c2 \u03b1\u03bd\u03ac \u03ba\u03bb\u03ae\u03c3\u03b7: \u03b1\u03c5\u03be\u03ac\u03bd\u03b5\u03b9 \u03ba\u03b1\u03b9 \u03c4\u03bf context \u03c0\u03bf\u03c5 \u03b5\u03c0\u03b1\u03bd\u03b5\u03c0\u03b5\u03be\u03b5\u03c1\u03b3\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c4\u03b1 \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b1 \u03b2\u03ae\u03bc\u03b1\u03c4\u03b1. \u039f\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03bf\u03c5\u03bd \u03b1\u03c5\u03c4\u03ae \u03c4\u03b7 \u03c3\u03c5\u03c3\u03c3\u03ce\u03c1\u03b5\u03c5\u03c3\u03b7 \u03c9\u03c2 reasoning residue.<\/p>\n<p>\u03a5\u03c0\u03ac\u03c1\u03c7\u03b5\u03b9 \u03c3\u03b7\u03bc\u03b1\u03bd\u03c4\u03b9\u03ba\u03ae \u03b4\u03b9\u03ac\u03ba\u03c1\u03b9\u03c3\u03b7 \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03b5 APIs \u03c0\u03bf\u03c5 \u03ba\u03c1\u03cd\u03b2\u03bf\u03c5\u03bd \u03c4\u03b7\u03bd \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03ae \u03c3\u03ba\u03ad\u03c8\u03b7 \u03ba\u03b1\u03b9 \u03c3\u03b5 workflows \u03c0\u03bf\u03c5 \u03bc\u03b5\u03c4\u03b1\u03c6\u03ad\u03c1\u03bf\u03c5\u03bd \u03c1\u03b7\u03c4\u03ac \u03b5\u03bd\u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03bf \u03ba\u03b5\u03af\u03bc\u03b5\u03bd\u03bf. \u03a3\u03c4\u03b7\u03bd \u03c0\u03c1\u03ce\u03c4\u03b7 \u03c0\u03b5\u03c1\u03af\u03c0\u03c4\u03c9\u03c3\u03b7, \u03c5\u03c8\u03b7\u03bb\u03cc RCS \u03b1\u03c5\u03be\u03ac\u03bd\u03b5\u03b9 \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03ba\u03b1\u03b9 latency \u03b1\u03bb\u03bb\u03ac \u03cc\u03c7\u03b9 \u03b1\u03bd\u03b1\u03b3\u03ba\u03b1\u03c3\u03c4\u03b9\u03ba\u03ac \u03c4\u03bf context. \u03a3\u03c4\u03b7 \u03b4\u03b5\u03cd\u03c4\u03b5\u03c1\u03b7, verbose traces \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03bd \u03bd\u03b1 \u03b1\u03bd\u03c4\u03b1\u03b3\u03c9\u03bd\u03b9\u03c3\u03c4\u03bf\u03cd\u03bd \u03bf\u03b4\u03b7\u03b3\u03af\u03b5\u03c2, retrieved evidence, tool outputs \u03ba\u03b1\u03b9 state \u03b3\u03b9\u03b1 \u03c7\u03ce\u03c1\u03bf \u03c3\u03c4\u03bf context window. \u0388\u03bd\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c0\u03bf\u03c5 \u03c6\u03b1\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 \u03bb\u03bf\u03b3\u03b9\u03ba\u03cc \u03c3\u03b5 single-turn benchmark \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03ad\u03c4\u03c3\u03b9 \u03bd\u03b1 \u03b3\u03af\u03bd\u03b5\u03b9 \u03b1\u03bd\u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03b5\u03c3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03c3\u03b5 \u03bc\u03b5\u03b3\u03ac\u03bb\u03b7 \u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03af\u03b1.<\/p>\n<p>\u0397 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03ae \u03b1\u03c1\u03c7\u03ae \u03b5\u03af\u03bd\u03b1\u03b9 \u03bd\u03b1 \u03b1\u03bd\u03c4\u03b9\u03c3\u03c4\u03bf\u03b9\u03c7\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 reasoning \u03c3\u03b5 step type \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03c3\u03b5 \u03bf\u03bb\u03cc\u03ba\u03bb\u03b7\u03c1\u03bf pipeline. \u0388\u03bd\u03b1\u03c2 orchestrator \u03b4\u03b5\u03bd \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c4\u03bf \u03b1\u03ba\u03c1\u03b9\u03b2\u03cc\u03c4\u03b5\u03c1\u03bf reasoning \u03c3\u03b5 routing \u03ae formatting. \u039f\u03b9 subagents \u03b4\u03b5\u03bd \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03bd\u03b1 \u03b5\u03c0\u03b9\u03c3\u03c4\u03c1\u03ad\u03c6\u03bf\u03c5\u03bd \u03cc\u03bb\u03b7 \u03c4\u03b7 \u03b4\u03b9\u03b1\u03b4\u03c1\u03bf\u03bc\u03ae \u03c3\u03ba\u03ad\u03c8\u03b7\u03c2 \u03cc\u03c4\u03b1\u03bd \u03b1\u03c1\u03ba\u03b5\u03af \u03ad\u03bd\u03b1 \u03c3\u03c5\u03bc\u03c0\u03b1\u03b3\u03ad\u03c2 \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1, evidence \u03ba\u03b1\u03b9 \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03bf state. \u03a4\u03bf context window \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03bd\u03c4\u03b9\u03bc\u03b5\u03c4\u03c9\u03c0\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b1\u03bd \u03ba\u03bf\u03b9\u03bd\u03cc\u03c7\u03c1\u03b7\u03c3\u03c4\u03bf\u03c2 \u03c0\u03c1\u03bf\u03cb\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03cc\u03c2.<\/p>\n<h2 id=\"plano-axiologisis\">\u0388\u03bd\u03b1 \u03c0\u03bb\u03ac\u03bd\u03bf \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7\u03c2 \u03b3\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03ae\u03c3\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 e-commerce teams<\/h2>\n<p>\u0397 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b3\u03af\u03bd\u03b5\u03b9 \u03c0\u03ac\u03bd\u03c9 \u03c3\u03b5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac tasks \u03ba\u03b1\u03b9 \u03bd\u03b1 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c4\u03b1\u03c5\u03c4\u03cc\u03c7\u03c1\u03bf\u03bd\u03b1 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1, generated tokens, latency, \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03ba\u03b1\u03b9 \u03b1\u03bd\u03ac\u03b3\u03ba\u03b7 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf\u03c5 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5. \u03a4\u03bf \u03c0\u03b1\u03c1\u03b1\u03ba\u03ac\u03c4\u03c9 workflow \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03c4\u03b7 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7 \u03c3\u03b5 \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03ae\u03c8\u03b9\u03bc\u03b7 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03b4\u03bf\u03ba\u03b9\u03bc\u03ae.<\/p>\n<div class=\"td-step-list\">\n<p class=\"td-step-list-title\">\u0395\u03c0\u03c4\u03ac \u03b2\u03ae\u03bc\u03b1\u03c4\u03b1 \u03b3\u03b9\u03b1 \u03bd\u03b1 \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03b5\u03c4\u03b5 \u03c4\u03bf reasoning tax<\/p>\n<ol>\n<li><span class=\"td-step-kicker\">Workload<\/span><strong>\u03a4\u03b1\u03be\u03b9\u03bd\u03bf\u03bc\u03ae\u03c3\u03c4\u03b5 \u03c4\u03b9\u03c2 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b5\u03c2<\/strong>\n<p>\u03a7\u03c9\u03c1\u03af\u03c3\u03c4\u03b5 tickets, \u03c0\u03c1\u03bf\u03b2\u03bb\u03ae\u03bc\u03b1\u03c4\u03b1 \u03ba\u03b1\u03c4\u03b1\u03bb\u03cc\u03b3\u03bf\u03c5, analytics \u03b5\u03c1\u03c9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1, campaign briefs \u03ae coding tasks \u03c3\u03b5 retrieval, transformation, sequential inference, generation \u03ba\u03b1\u03b9 verification.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Validation<\/span><strong>\u03a7\u03c4\u03af\u03c3\u03c4\u03b5 \u03b1\u03bd\u03c4\u03b9\u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03b5\u03c5\u03c4\u03b9\u03ba\u03cc \u03c3\u03cd\u03bd\u03bf\u03bb\u03bf \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5<\/strong>\n<p>\u03a7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03ae\u03c3\u03c4\u03b5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03c0\u03b5\u03c1\u03b9\u03c0\u03c4\u03ce\u03c3\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 edge cases \u03c7\u03c9\u03c1\u03af\u03c2 \u03c0\u03b5\u03c1\u03b9\u03c4\u03c4\u03ac \u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03b9\u03ba\u03ac \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1. \u03a4\u03b1 \u03b3\u03b5\u03bd\u03b9\u03ba\u03ac benchmarks \u03b4\u03b5\u03bd \u03c5\u03c0\u03bf\u03ba\u03b1\u03b8\u03b9\u03c3\u03c4\u03bf\u03cd\u03bd \u03c4\u03b7 \u03b4\u03b9\u03ba\u03ae \u03c3\u03b1\u03c2 \u03ba\u03b1\u03c4\u03b1\u03bd\u03bf\u03bc\u03ae \u03b5\u03c1\u03b3\u03b1\u03c3\u03b9\u03ce\u03bd.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Baseline<\/span><strong>\u039a\u03bb\u03b5\u03b9\u03b4\u03ce\u03c3\u03c4\u03b5 \u03ad\u03bd\u03b1 non-reasoning \u03c3\u03b7\u03bc\u03b5\u03af\u03bf \u03b1\u03bd\u03b1\u03c6\u03bf\u03c1\u03ac\u03c2<\/strong>\n<p>\u039a\u03c1\u03b1\u03c4\u03ae\u03c3\u03c4\u03b5 \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03ac prompt, tools, retrieval \u03ba\u03b1\u03b9 evaluation rules, \u03ce\u03c3\u03c4\u03b5 \u03b7 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03bd\u03b1 \u03bc\u03b7\u03bd \u03c0\u03c1\u03bf\u03ad\u03c1\u03c7\u03b5\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03ac\u03bb\u03bb\u03b7 \u03b1\u03bb\u03bb\u03b1\u03b3\u03ae \u03c3\u03c4\u03bf pipeline.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Effort<\/span><strong>\u03a3\u03c5\u03b3\u03ba\u03c1\u03af\u03bd\u03b5\u03c4\u03b5 medium, high \u03ba\u03b1\u03b9 \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b7 \u03b2\u03b1\u03b8\u03bc\u03af\u03b4\u03b1<\/strong>\n<p>\u039a\u03b1\u03c4\u03b1\u03b3\u03c1\u03ac\u03c8\u03c4\u03b5 accuracy \u03ae task completion \u03bc\u03b1\u03b6\u03af \u03bc\u03b5 generated tokens, latency \u03ba\u03b1\u03b9 \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2. \u039c\u03b7\u03bd \u03be\u03b5\u03ba\u03b9\u03bd\u03ac\u03c4\u03b5 \u03bc\u03b7\u03c7\u03b1\u03bd\u03b9\u03ba\u03ac \u03b1\u03c0\u03cc maximum effort.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Review<\/span><strong>\u039c\u03b5\u03c4\u03c1\u03ae\u03c3\u03c4\u03b5 \u03c4\u03b7\u03bd \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7 \u03c0\u03b1\u03c1\u03ad\u03bc\u03b2\u03b1\u03c3\u03b7<\/strong>\n<p>\u03a4\u03bf business metric \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 human-review rate \u03ae \u03b4\u03b9\u03bf\u03c1\u03b8\u03ce\u03c3\u03b5\u03b9\u03c2 \u03b1\u03bd\u03ac task. \u0395\u03bb\u03ad\u03b3\u03be\u03c4\u03b5 \u03be\u03b5\u03c7\u03c9\u03c1\u03b9\u03c3\u03c4\u03ac \u03c4\u03b7\u03bd \u03b1\u03be\u03b9\u03bf\u03c0\u03b9\u03c3\u03c4\u03af\u03b1 \u03cc\u03c4\u03b1\u03bd \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 <a href=\"https:\/\/twodots.gr\/llm-judges-ai-agents-megalyteros-kritis-den-arkei\/\">LLM judges \u03b3\u03b9\u03b1 \u03c4\u03b7\u03bd \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 AI agents<\/a>.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Routing<\/span><strong>\u0394\u03ce\u03c3\u03c4\u03b5 reasoning \u03bc\u03cc\u03bd\u03bf \u03c3\u03c4\u03b1 \u03b2\u03ae\u03bc\u03b1\u03c4\u03b1 \u03c0\u03bf\u03c5 \u03c4\u03bf \u03ba\u03b5\u03c1\u03b4\u03af\u03b6\u03bf\u03c5\u03bd<\/strong>\n<p>\u03a3\u03c4\u03b5\u03af\u03bb\u03c4\u03b5 planning, computation \u03ba\u03b1\u03b9 verification \u03c3\u03c4\u03bf reasoning \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf, \u03b5\u03bd\u03ce routine retrieval, classification \u03ba\u03b1\u03b9 formatting \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03bd \u03bd\u03b1 \u03bc\u03ad\u03bd\u03bf\u03c5\u03bd \u03c3\u03c4\u03bf \u03b3\u03c1\u03b7\u03b3\u03bf\u03c1\u03cc\u03c4\u03b5\u03c1\u03bf baseline.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Pra\u0107enje<\/span><strong>\u0395\u03c0\u03b1\u03bd\u03b5\u03bb\u03ad\u03b3\u03be\u03c4\u03b5 \u03bc\u03b5\u03c4\u03ac \u03b1\u03c0\u03cc \u03ba\u03ac\u03b8\u03b5 \u03bf\u03c5\u03c3\u03b9\u03b1\u03c3\u03c4\u03b9\u03ba\u03ae \u03b1\u03bb\u03bb\u03b1\u03b3\u03ae<\/strong>\n<p>\u03a4\u03bf mix \u03b5\u03c1\u03b3\u03b1\u03c3\u03b9\u03ce\u03bd, \u03c4\u03b1 prompts, \u03bf\u03b9 \u03c4\u03b9\u03bc\u03ad\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03b1\u03bb\u03bb\u03ac\u03b6\u03bf\u03c5\u03bd, \u03cc\u03c0\u03c9\u03c2 \u03b1\u03bb\u03bb\u03ac\u03b6\u03bf\u03c5\u03bd \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03c0\u03bf\u03c5 \u03c4\u03c1\u03bf\u03c6\u03bf\u03b4\u03bf\u03c4\u03bf\u03cd\u03bd \u03c4\u03b9\u03c2 <a href=\"https:\/\/twodots.gr\/analytics-platforms-data-decisions\/\">analytics \u03c0\u03bb\u03b1\u03c4\u03c6\u03cc\u03c1\u03bc\u03b5\u03c2 \u03b3\u03b9\u03b1 \u03b1\u03c0\u03bf\u03c6\u03ac\u03c3\u03b5\u03b9\u03c2 growth<\/a>. \u0397 \u03b1\u03c0\u03bf\u03b4\u03bf\u03c4\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03c4\u03bf\u03c5 \u03c0\u03c1\u03bf\u03b7\u03b3\u03bf\u03cd\u03bc\u03b5\u03bd\u03bf\u03c5 \u03bc\u03ae\u03bd\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03bc\u03cc\u03bd\u03b9\u03bc\u03b7 \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae.<\/p>\n<\/li>\n<\/ol>\n<\/div>\n<h2 id=\"anagnosi-arithmon\">\u03a0\u03ce\u03c2 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03b2\u03ac\u03c3\u03b5\u03c4\u03b5 \u03c4\u03b1 \u03bd\u03bf\u03cd\u03bc\u03b5\u03c1\u03b1 \u03c7\u03c9\u03c1\u03af\u03c2 \u03bd\u03b1 \u03c4\u03b1 \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c8\u03b5\u03c4\u03b5 \u03c3\u03b5 \u03c5\u03c0\u03cc\u03c3\u03c7\u03b5\u03c3\u03b7<\/h2>\n<p>\u03a4\u03b1 151 runs \u03ba\u03b1\u03b9 \u03bf\u03b9 77 reasoning\/non-reasoning \u03c3\u03c5\u03b3\u03ba\u03c1\u03af\u03c3\u03b5\u03b9\u03c2 \u03b4\u03af\u03bd\u03bf\u03c5\u03bd \u03b9\u03c3\u03c7\u03c5\u03c1\u03cc\u03c4\u03b5\u03c1\u03bf \u03c3\u03ae\u03bc\u03b1 \u03b1\u03c0\u03cc \u03ad\u03bd\u03b1 \u03bc\u03b5\u03bc\u03bf\u03bd\u03c9\u03bc\u03ad\u03bd\u03bf demo, \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03bf\u03cd\u03bd \u03b5\u03b3\u03b3\u03cd\u03b7\u03c3\u03b7 \u03b3\u03b9\u03b1 \u03ba\u03ac\u03b8\u03b5 \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03b9\u03ba\u03cc \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1. \u03a4\u03b1 benchmarks \u03ad\u03c7\u03bf\u03c5\u03bd \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b7 \u03b4\u03b9\u03b1\u03c4\u03cd\u03c0\u03c9\u03c3\u03b7, scoring \u03ba\u03b1\u03b9 \u03ba\u03b1\u03c4\u03b1\u03bd\u03bf\u03bc\u03ae \u03b5\u03c1\u03c9\u03c4\u03ae\u03c3\u03b5\u03c9\u03bd. \u0388\u03bd\u03b1 <a href=\"https:\/\/twodots.gr\/vasika-charaktiristika-platformas-ilektronikou-emporiou\/\">e-commerce workflow \u03bc\u03ad\u03c3\u03b1 \u03c3\u03b5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03c0\u03bb\u03b1\u03c4\u03c6\u03cc\u03c1\u03bc\u03b1 \u03b7\u03bb\u03b5\u03ba\u03c4\u03c1\u03bf\u03bd\u03b9\u03ba\u03bf\u03cd \u03b5\u03bc\u03c0\u03bf\u03c1\u03af\u03bf\u03c5<\/a>, \u03bc\u03b5 \u03b5\u03bb\u03bb\u03b7\u03bd\u03b9\u03ba\u03cc \u03ba\u03b1\u03c4\u03ac\u03bb\u03bf\u03b3\u03bf, \u03b5\u03c1\u03b3\u03b1\u03bb\u03b5\u03af\u03b1 \u03b1\u03bd\u03b1\u03b6\u03ae\u03c4\u03b7\u03c3\u03b7\u03c2 \u03ba\u03b1\u03b9 \u03b1\u03c5\u03c3\u03c4\u03b7\u03c1\u03bf\u03cd\u03c2 \u03ba\u03b1\u03bd\u03cc\u03bd\u03b5\u03c2 brand, \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c3\u03c5\u03bc\u03c0\u03b5\u03c1\u03b9\u03c6\u03b5\u03c1\u03b8\u03b5\u03af \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac \u03b1\u03c0\u03cc \u03c4\u03bf MMLU-Pro \u03ae \u03c4\u03bf LiveCodeBench. \u0397 \u03b1\u03be\u03af\u03b1 \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2 \u03b2\u03c1\u03af\u03c3\u03ba\u03b5\u03c4\u03b1\u03b9 \u03ba\u03c5\u03c1\u03af\u03c9\u03c2 \u03c3\u03c4\u03b7 \u03bc\u03b5\u03b8\u03bf\u03b4\u03bf\u03bb\u03bf\u03b3\u03af\u03b1 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7\u03c2 \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b1 \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b1 patterns, \u03cc\u03c7\u03b9 \u03c3\u03b5 \u03bc\u03b9\u03b1 \u03ad\u03c4\u03bf\u03b9\u03bc\u03b7 \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae vendor.<\/p>\n<p>\u03a7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b5\u03c0\u03af\u03c3\u03b7\u03c2 \u03b4\u03b9\u03ac\u03ba\u03c1\u03b9\u03c3\u03b7 \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03b5 token efficiency \u03ba\u03b1\u03b9 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03cc \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1. \u0388\u03bd\u03b1 \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc TES \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03c0\u03bf\u03b4\u03b5\u03ba\u03c4\u03cc \u03cc\u03c4\u03b1\u03bd \u03ad\u03bd\u03b1 \u03c3\u03c0\u03ac\u03bd\u03b9\u03bf \u03bb\u03ac\u03b8\u03bf\u03c2 \u03ad\u03c7\u03b5\u03b9 \u03bc\u03b5\u03b3\u03ac\u03bb\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2, \u03b1\u03c1\u03ba\u03b5\u03af \u03c4\u03bf reasoning \u03bd\u03b1 \u03bc\u03b5\u03b9\u03ce\u03bd\u03b5\u03b9 \u03b1\u03c0\u03bf\u03b4\u03b5\u03b4\u03b5\u03b9\u03b3\u03bc\u03ad\u03bd\u03b1 \u03b1\u03c5\u03c4\u03cc\u03bd \u03c4\u03bf\u03bd \u03ba\u03af\u03bd\u03b4\u03c5\u03bd\u03bf. \u0391\u03bd\u03c4\u03af\u03c3\u03c4\u03c1\u03bf\u03c6\u03b1, \u03ad\u03bd\u03b1 \u03c5\u03c8\u03b7\u03bb\u03cc TES \u03b4\u03b5\u03bd \u03b1\u03c1\u03ba\u03b5\u03af \u03cc\u03c4\u03b1\u03bd \u03b7 \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03b5\u03b9 \u03ba\u03ac\u03c4\u03c9 \u03b1\u03c0\u03cc \u03c4\u03bf \u03b1\u03c0\u03b1\u03b9\u03c4\u03bf\u03cd\u03bc\u03b5\u03bd\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf \u03ae \u03cc\u03c4\u03b1\u03bd \u03c4\u03bf latency \u03ba\u03b1\u03c4\u03b1\u03c3\u03c4\u03c1\u03ad\u03c6\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b5\u03bc\u03c0\u03b5\u03b9\u03c1\u03af\u03b1 \u03c0\u03b5\u03bb\u03ac\u03c4\u03b7. \u0397 \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03b1\u03c0\u03cc\u03c6\u03b1\u03c3\u03b7 \u03c3\u03c5\u03bd\u03b4\u03c5\u03ac\u03b6\u03b5\u03b9 quality threshold, \u03cc\u03b3\u03ba\u03bf, \u03c7\u03c1\u03cc\u03bd\u03bf \u03b1\u03c0\u03cc\u03ba\u03c1\u03b9\u03c3\u03b7\u03c2, \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf\u03c5 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5 \u03ba\u03b1\u03b9 \u03c3\u03c5\u03bd\u03ad\u03c0\u03b5\u03b9\u03b5\u03c2 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b1\u03c2.<\/p>\n<p>\u0393\u03b9\u03b1 \u03bd\u03b1 \u03b1\u03c0\u03bf\u03c6\u03cd\u03b3\u03b5\u03b9 \u03ad\u03bd\u03b1 team \u03c4\u03b7 \u03bb\u03ac\u03b8\u03bf\u03c2 \u03b3\u03b5\u03bd\u03af\u03ba\u03b5\u03c5\u03c3\u03b7, \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03ba\u03c1\u03b1\u03c4\u03ac \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03ac prompt, tools \u03ba\u03b1\u03b9 evaluation rules \u03ba\u03b1\u03c4\u03ac \u03c4\u03b7 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7 effort levels. \u039d\u03b1 \u03b1\u03bd\u03b1\u03c6\u03ad\u03c1\u03b5\u03b9 \u03bc\u03b1\u03b6\u03af accuracy, generated tokens \u03ba\u03b1\u03b9 confidence intervals \u03ae \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03ae\u03c8\u03b5\u03b9\u03c2 \u03cc\u03c0\u03bf\u03c5 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b5\u03c6\u03b9\u03ba\u03c4\u03cc. \u039d\u03b1 \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 outliers \u03b1\u03bd\u03c4\u03af \u03bd\u03b1 \u03c4\u03b1 \u03ba\u03c1\u03cd\u03b2\u03b5\u03b9 \u03c3\u03c4\u03bf\u03bd \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf. \u039a\u03b1\u03b9 \u03bd\u03b1 \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03ac\u03bd\u03b5\u03b9 \u03c4\u03b7 \u03b4\u03bf\u03ba\u03b9\u03bc\u03ae \u03bc\u03b5\u03c4\u03ac \u03b1\u03c0\u03cc \u03b1\u03bb\u03bb\u03b1\u03b3\u03ae \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5, \u03c4\u03b9\u03bc\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7\u03c2 \u03ae agent architecture. \u0388\u03c4\u03c3\u03b9, \u03c4\u03bf TES \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03b5\u03af \u03c9\u03c2 \u03b6\u03c9\u03bd\u03c4\u03b1\u03bd\u03cc operational metric \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03c9\u03c2 \u03c3\u03c4\u03b1\u03c4\u03b9\u03ba\u03cc\u03c2 \u03b1\u03c1\u03b9\u03b8\u03bc\u03cc\u03c2 \u03b1\u03c0\u03cc \u03ad\u03bd\u03b1 paper.<\/p>\n<h2 id=\"periorismoi\">\u03a0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03af \u03c0\u03bf\u03c5 \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03ba\u03c1\u03c5\u03c6\u03c4\u03bf\u03cd\u03bd<\/h2>\n<p>\u0397 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03bc\u03ad\u03c3\u03bf\u03c5\u03c2 token counts \u03b1\u03bd\u03ac benchmark, \u03bf\u03b9 \u03bf\u03c0\u03bf\u03af\u03bf\u03b9 \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03bd \u03bd\u03b1 \u03b5\u03c0\u03b7\u03c1\u03b5\u03b1\u03c3\u03c4\u03bf\u03cd\u03bd \u03b1\u03c0\u03cc outliers. \u039c\u03af\u03b1 \u03b6\u03b5\u03cd\u03be\u03b7 IFBench \u03b5\u03be\u03b1\u03b9\u03c1\u03ad\u03b8\u03b7\u03ba\u03b5 \u03b1\u03c0\u03cc \u03c4\u03b1 aggregate statistics \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03b7 non-reasoning \u03b5\u03ba\u03b4\u03bf\u03c7\u03ae \u03c4\u03bf\u03c5 Grok 4.20 0309 \u03c0\u03b1\u03c1\u03ae\u03b3\u03b1\u03b3\u03b5 870.000 output tokens \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 48.000 \u03c4\u03b7\u03c2 reasoning \u03b5\u03ba\u03b4\u03bf\u03c7\u03ae\u03c2, \u03c3\u03c5\u03bc\u03c0\u03b5\u03c1\u03b9\u03c6\u03bf\u03c1\u03ac \u03c0\u03bf\u03c5 \u03bf\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03b1\u03c0\u03bf\u03b4\u03af\u03b4\u03bf\u03c5\u03bd \u03c0\u03b9\u03b8\u03b1\u03bd\u03ce\u03c2 \u03c3\u03b5 failure mode \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03c3\u03b5 \u03bf\u03c5\u03c3\u03b9\u03b1\u03c3\u03c4\u03b9\u03ba\u03cc baseline. \u03a4\u03bf record \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03b5\u03b9 \u03c3\u03c4\u03bf \u03b4\u03b7\u03bc\u03cc\u03c3\u03b9\u03bf repository.<\/p>\n<p>\u039f\u03b9 \u03c4\u03b9\u03bc\u03ad\u03c2 RCS, cloud cost \u03ba\u03b1\u03b9 DCM \u03b2\u03b1\u03c3\u03af\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c3\u03b5 API pricing \u03c3\u03c4\u03b1 \u03c4\u03ad\u03bb\u03b7 \u039c\u03b1\u0390\u03bf\u03c5 2026 \u03ba\u03b1\u03b9 \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03bc\u03b5\u03c4\u03b1\u03c6\u03ad\u03c1\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b1\u03c5\u03c4\u03bf\u03cd\u03c3\u03b9\u03b5\u03c2 \u03c3\u03b5 \u03bc\u03b5\u03bb\u03bb\u03bf\u03bd\u03c4\u03b9\u03ba\u03cc budget. \u03a4\u03bf TES-A \u03c3\u03c5\u03b3\u03ba\u03c1\u03af\u03bd\u03b5\u03b9 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ad\u03c2 \u03bf\u03b9\u03ba\u03bf\u03b3\u03ad\u03bd\u03b5\u03b9\u03b5\u03c2, \u03b5\u03bd\u03ce \u03c4\u03bf on-premises \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c5\u03c0\u03bf\u03b8\u03ad\u03c4\u03b5\u03b9 \u03b1\u03be\u03b9\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03b9\u03ba\u03b1\u03bd\u03ae \u03bd\u03b1 \u03b1\u03c0\u03bf\u03c3\u03b2\u03ad\u03c3\u03b5\u03b9 \u03c4\u03b7\u03bd \u03c5\u03c0\u03bf\u03b4\u03bf\u03bc\u03ae. \u03a4\u03ad\u03bb\u03bf\u03c2, \u03c4\u03bf single-turn TES \u03b4\u03b5\u03bd \u03bc\u03b5\u03c4\u03c1\u03ac \u03b1\u03ba\u03cc\u03bc\u03b7 \u03c0\u03bb\u03ae\u03c1\u03c9\u03c2 \u03c4\u03b7\u03bd \u03b1\u03cd\u03be\u03b7\u03c3\u03b7 context \u03ba\u03b1\u03b9 \u03c4\u03b7\u03bd end-to-end \u03b5\u03c0\u03b9\u03c4\u03c5\u03c7\u03af\u03b1 \u03b5\u03bd\u03cc\u03c2 agentic pipeline. \u039f\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03c0\u03c1\u03bf\u03c4\u03b5\u03af\u03bd\u03bf\u03c5\u03bd \u03bc\u03b5\u03bb\u03bb\u03bf\u03bd\u03c4\u03b9\u03ba\u03cc agentic TES \u03bc\u03b5 \u03bb\u03bf\u03b3\u03b9\u03c3\u03c4\u03b9\u03ba\u03ae \u03b1\u03bd\u03ac \u03b2\u03ae\u03bc\u03b1.<\/p>\n<h2 id=\"symperasma\">\u0397 \u03bf\u03c5\u03c3\u03af\u03b1 \u03b3\u03b9\u03b1 \u03bc\u03b9\u03b1 \u03b4\u03b9\u03bf\u03af\u03ba\u03b7\u03c3\u03b7 \u03c0\u03bf\u03c5 \u03b5\u03c0\u03b5\u03bd\u03b4\u03cd\u03b5\u03b9 \u03c3\u03b5 AI<\/h2>\n<p>\u03a4\u03bf reasoning \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 premium \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03c0\u03bf\u03c5 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03bd\u03b5\u03c1\u03b3\u03bf\u03c0\u03bf\u03b9\u03b5\u03af\u03c4\u03b1\u03b9 \u03bc\u03b7\u03c7\u03b1\u03bd\u03b9\u03ba\u03ac. \u0395\u03af\u03bd\u03b1\u03b9 \u03ad\u03bd\u03b1\u03c2 \u03c0\u03cc\u03c1\u03bf\u03c2 \u03c0\u03bf\u03c5 \u03b1\u03b3\u03bf\u03c1\u03ac\u03b6\u03b5\u03b9 \u03c0\u03b9\u03b8\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b7\u03c2 \u03b1\u03c0\u03cc\u03c6\u03b1\u03c3\u03b7\u03c2 \u03c3\u03b5 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf\u03c5\u03c2 \u03c4\u03cd\u03c0\u03bf\u03c5\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2. \u0397 \u03ad\u03c1\u03b5\u03c5\u03bd\u03b1 \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03bf\u03b9 \u03b1\u03bb\u03bb\u03b7\u03bb\u03b5\u03be\u03b1\u03c1\u03c4\u03ce\u03bc\u03b5\u03bd\u03b5\u03c2 \u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03af\u03b5\u03c2 \u03bb\u03bf\u03b3\u03b9\u03ba\u03ae\u03c2 \u03ba\u03b5\u03c1\u03b4\u03af\u03b6\u03bf\u03c5\u03bd \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf, \u03b7 \u03b1\u03bd\u03ac\u03ba\u03bb\u03b7\u03c3\u03b7 \u03b3\u03bd\u03ce\u03c3\u03b7\u03c2 \u03bb\u03b9\u03b3\u03cc\u03c4\u03b5\u03c1\u03bf \u03ba\u03b1\u03b9 \u03c4\u03bf maximum effort \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03bc\u03b5\u03b9\u03ce\u03c3\u03b5\u03b9 \u03b1\u03ba\u03cc\u03bc\u03b7 \u03ba\u03b1\u03b9 \u03c4\u03b7\u03bd \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1.<\/p>\n<p>\u0393\u03b9\u03b1 marketers, e-commerce owners \u03ba\u03b1\u03b9 AI product teams, \u03b7 \u03c3\u03c9\u03c3\u03c4\u03ae \u03b5\u03c1\u03ce\u03c4\u03b7\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9: \u00ab\u03a3\u03b5 \u03c0\u03bf\u03b9\u03bf \u03b2\u03ae\u03bc\u03b1 \u03b7 \u03c0\u03c1\u03cc\u03c3\u03b8\u03b5\u03c4\u03b7 \u03c3\u03ba\u03ad\u03c8\u03b7 \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03ad\u03bd\u03b1 \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03b9\u03bc\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1;\u00bb. \u0391\u03bd \u03b4\u03b5\u03bd \u03c5\u03c0\u03ac\u03c1\u03c7\u03b5\u03b9 validation set, token accounting \u03ba\u03b1\u03b9 \u03c3\u03b1\u03c6\u03ad\u03c2 quality threshold, \u03c4\u03bf reasoning tax \u03bc\u03ad\u03bd\u03b5\u03b9 \u03b1\u03cc\u03c1\u03b1\u03c4\u03bf. \u038c\u03c4\u03b1\u03bd \u03b1\u03c5\u03c4\u03ac \u03c5\u03c0\u03ac\u03c1\u03c7\u03bf\u03c5\u03bd, \u03c4\u03bf AI stack \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b3\u03af\u03bd\u03b5\u03b9 \u03c4\u03b1\u03c5\u03c4\u03cc\u03c7\u03c1\u03bf\u03bd\u03b1 \u03c0\u03b9\u03bf \u03b1\u03ba\u03c1\u03b9\u03b2\u03ad\u03c2, \u03c0\u03b9\u03bf \u03c0\u03c1\u03bf\u03b2\u03bb\u03ad\u03c8\u03b9\u03bc\u03bf \u03ba\u03b1\u03b9 \u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac \u03b5\u03bb\u03b5\u03b3\u03c7\u03cc\u03bc\u03b5\u03bd\u03bf.<\/p>\n<section class=\"td-service-cta\">\n<div class=\"td-service-cta-content\">\n<p class=\"td-service-cta-eyebrow\">Poslovna automatizacija i umjetna inteligencija<\/p>\n<p class=\"td-service-cta-title\"><strong>\u039c\u03b5\u03c4\u03c1\u03ae\u03c3\u03c4\u03b5 \u03c4\u03b7\u03bd \u03b1\u03be\u03af\u03b1 \u03c4\u03bf\u03c5 reasoning \u03c0\u03c1\u03b9\u03bd \u03c4\u03bf \u03b2\u03ac\u03bb\u03b5\u03c4\u03b5 \u03c0\u03b1\u03bd\u03c4\u03bf\u03cd<\/strong><\/p>\n<p>\u0397 TWO DOTS \u03c7\u03b1\u03c1\u03c4\u03bf\u03b3\u03c1\u03b1\u03c6\u03b5\u03af \u03c4\u03b1 \u03b2\u03ae\u03bc\u03b1\u03c4\u03b1 \u03b5\u03bd\u03cc\u03c2 AI workflow, \u03bf\u03c1\u03af\u03b6\u03b5\u03b9 validation set \u03ba\u03b1\u03b9 quality thresholds \u03ba\u03b1\u03b9 \u03c3\u03c7\u03b5\u03b4\u03b9\u03ac\u03b6\u03b5\u03b9 routing, \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b1 gates \u03ba\u03b1\u03b9 monitoring \u03ce\u03c3\u03c4\u03b5 \u03c4\u03bf \u03b5\u03c0\u03b9\u03c0\u03bb\u03ad\u03bf\u03bd reasoning \u03bd\u03b1 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af\u03c4\u03b1\u03b9 \u03bc\u03cc\u03bd\u03bf \u03cc\u03c0\u03bf\u03c5 \u03b2\u03b5\u03bb\u03c4\u03b9\u03ce\u03bd\u03b5\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03b9\u03bc\u03b1 \u03c4\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1.<\/p>\n<div class=\"td-service-cta-actions\"><a class=\"td-service-cta-button\" href=\"https:\/\/twodots.gr\/aftomatismoi-epicheiriseon-ai\/\">Pogledajte automatizaciju poslovanja i umjetnu inteligenciju<\/a><\/div>\n<\/div>\n<\/section>\n<section id=\"sychnes-erotiseis\" class=\"td-faq-section\">\n<div class=\"td-faq\">\n<p class=\"td-faq-heading\">\u010cesto postavljana pitanja<\/p>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a4\u03b9 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf reasoning tax \u03c4\u03c9\u03bd LLMs;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0395\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf \u03c0\u03c1\u03cc\u03c3\u03b8\u03b5\u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03c3\u03b5 generated tokens, latency \u03ba\u03b1\u03b9 \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03c4\u03b9\u03ba\u03bf\u03cd\u03c2 \u03c0\u03cc\u03c1\u03bf\u03c5\u03c2 \u03c0\u03bf\u03c5 \u03c3\u03c5\u03bd\u03bf\u03b4\u03b5\u03cd\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b5\u03ba\u03c4\u03b5\u03c4\u03b1\u03bc\u03ad\u03bd\u03b7 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03ae \u03b5\u03c0\u03b5\u03be\u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03c0\u03c1\u03b9\u03bd \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7. \u0391\u03be\u03b9\u03bf\u03bb\u03bf\u03b3\u03b5\u03af\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03c3\u03c7\u03ad\u03c3\u03b7 \u03bc\u03b5 \u03c4\u03bf \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03ba\u03ad\u03c1\u03b4\u03bf\u03c2 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 \u03c3\u03c4\u03bf \u03af\u03b4\u03b9\u03bf task.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a4\u03b9 \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03ad\u03bd\u03b1 TES \u03c0\u03ac\u03bd\u03c9 \u03b1\u03c0\u03cc 1;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u039c\u03b5 \u03c4\u03b7 \u03c3\u03cd\u03bc\u03b2\u03b1\u03c3\u03b7 \u03c4\u03b7\u03c2 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7\u03c2, \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03bf\u03b9 \u03c0\u03bf\u03c3\u03bf\u03c3\u03c4\u03b9\u03b1\u03af\u03b5\u03c2 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b5\u03c2 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2 \u03c0\u03bf\u03c5 \u03ba\u03b5\u03c1\u03b4\u03ae\u03b8\u03b7\u03ba\u03b1\u03bd \u03c5\u03c0\u03b5\u03c1\u03b2\u03b1\u03af\u03bd\u03bf\u03c5\u03bd \u03c4\u03bf\u03bd \u03c0\u03bf\u03bb\u03bb\u03b1\u03c0\u03bb\u03b1\u03c3\u03b9\u03b1\u03c3\u03c4\u03ae generated tokens. \u0394\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03b1\u03b8\u03bf\u03bb\u03b9\u03ba\u03cc ROI, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03ba\u03ac\u03b8\u03b5 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03b1\u03c0\u03bf\u03c4\u03b9\u03bc\u03ac \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac \u03c4\u03b7\u03bd \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a3\u03b5 \u03c0\u03bf\u03b9\u03b5\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b5\u03c2 \u03b1\u03c0\u03bf\u03b4\u03af\u03b4\u03b5\u03b9 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b1 \u03c4\u03bf reasoning;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u03a3\u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2 \u03b1\u03c0\u03bf\u03b4\u03af\u03b4\u03b5\u03b9 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b1 \u03c3\u03b5 sequential inference, \u03cc\u03c0\u03c9\u03c2 \u03bc\u03b1\u03b8\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b1\u03c0\u03bf\u03b4\u03b5\u03af\u03be\u03b5\u03b9\u03c2, \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae \u03ba\u03b1\u03b9 debugging \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1 \u03ae \u03c0\u03c1\u03bf\u03b2\u03bb\u03ae\u03bc\u03b1\u03c4\u03b1 \u03bc\u03b5 \u03b1\u03bb\u03bb\u03b7\u03bb\u03b5\u03be\u03b1\u03c1\u03c4\u03ce\u03bc\u03b5\u03bd\u03bf\u03c5\u03c2 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2. \u03a3\u03b5 recall-heavy \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b5\u03c2 \u03b7 \u03b1\u03c0\u03cc\u03b4\u03bf\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c3\u03c5\u03c7\u03bd\u03ac \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03b7.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03ad\u03bd\u03b1 \u03b4\u03cd\u03c3\u03ba\u03bf\u03bb\u03bf task \u03bd\u03b1 \u03c4\u03c1\u03ad\u03c7\u03b5\u03b9 \u03c0\u03ac\u03bd\u03c4\u03b1 \u03c3\u03b5 maximum effort;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u038c\u03c7\u03b9. \u0397 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7 \u03b2\u03c1\u03af\u03c3\u03ba\u03b5\u03b9 \u03c6\u03b8\u03af\u03bd\u03bf\u03c5\u03c3\u03b1 marginal \u03b1\u03c0\u03cc\u03b4\u03bf\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03c0\u03b5\u03c1\u03b9\u03c0\u03c4\u03ce\u03c3\u03b5\u03b9\u03c2 \u03c0\u03c4\u03ce\u03c3\u03b7\u03c2 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2 \u03c3\u03c4\u03bf \u03c5\u03c8\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf effort. \u0397 \u03b1\u03cd\u03be\u03b7\u03c3\u03b7 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 validation \u03c3\u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf task, \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03ba\u03b1\u03b9 workload.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a4\u03b9 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf Reasoning Cost Share;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0395\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf \u03c0\u03bf\u03c3\u03bf\u03c3\u03c4\u03cc \u03c4\u03bf\u03c5 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03bf\u03cd inference cost \u03c0\u03bf\u03c5 \u03b1\u03c0\u03bf\u03b4\u03af\u03b4\u03b5\u03c4\u03b1\u03b9 \u03c3\u03c4\u03b1 reasoning tokens. \u0397 \u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03b7 \u03c4\u03b9\u03bc\u03ae \u03c3\u03c4\u03b1 reasoning runs \u03c4\u03b7\u03c2 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7\u03c2 \u03ae\u03c4\u03b1\u03bd 94,7%, \u03b1\u03bb\u03bb\u03ac \u03c4\u03bf metric \u03b5\u03af\u03bd\u03b1\u03b9 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03b9\u03ba\u03cc \u03ba\u03b1\u03b9 \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03b5\u03af \u03bc\u03cc\u03bd\u03bf \u03c4\u03bf\u03c5 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae\u03c2.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a4\u03bf on-premises \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03c4\u03bf TES;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u038c\u03c7\u03b9, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03c4\u03bf TES \u03b2\u03b1\u03c3\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b5 token ratios. \u039c\u03c0\u03bf\u03c1\u03b5\u03af \u03cc\u03bc\u03c9\u03c2 \u03bd\u03b1 \u03b1\u03bb\u03bb\u03ac\u03be\u03b5\u03b9 \u03c4\u03bf \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03c4\u03b7\u03c2 \u03af\u03b4\u03b9\u03b1\u03c2 \u03b5\u03c0\u03af\u03b4\u03bf\u03c3\u03b7\u03c2. \u0397 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03c0\u03b5\u03c1\u03b9\u03bb\u03b1\u03bc\u03b2\u03ac\u03bd\u03b5\u03b9 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b1\u03be\u03b9\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7, hardware, \u03b5\u03bd\u03ad\u03c1\u03b3\u03b5\u03b9\u03b1, \u03c3\u03c5\u03bd\u03c4\u03ae\u03c1\u03b7\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03b4\u03b9\u03ac\u03c1\u03ba\u03b5\u03b9\u03b1 \u03b1\u03c0\u03cc\u03c3\u03b2\u03b5\u03c3\u03b7\u03c2.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03ce\u03c2 \u03b5\u03c6\u03b1\u03c1\u03bc\u03cc\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b7 \u03bb\u03bf\u03b3\u03b9\u03ba\u03ae \u03c3\u03b5 \u03ad\u03bd\u03b1\u03bd AI agent;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u039c\u03b5 reasoning \u03bc\u03cc\u03bd\u03bf \u03c3\u03c4\u03b1 \u03b2\u03ae\u03bc\u03b1\u03c4\u03b1 \u03c0\u03bf\u03c5 \u03b1\u03c0\u03b1\u03b9\u03c4\u03bf\u03cd\u03bd planning, computation \u03ae verification, \u03bc\u03b5 \u03c3\u03c5\u03bc\u03c0\u03b1\u03b3\u03ae handoffs \u03ba\u03b1\u03b9 \u03bc\u03b5 \u03bc\u03ad\u03c4\u03c1\u03b7\u03c3\u03b7 \u03c4\u03bf\u03c5 context \u03c0\u03bf\u03c5 \u03c3\u03c5\u03c3\u03c3\u03c9\u03c1\u03b5\u03cd\u03b5\u03c4\u03b1\u03b9. Routing \u03ba\u03b1\u03b9 formatting \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03bd \u03c3\u03c5\u03c7\u03bd\u03ac \u03bd\u03b1 \u03b5\u03ba\u03c4\u03b5\u03bb\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03c7\u03c9\u03c1\u03af\u03c2 \u03c5\u03c8\u03b7\u03bb\u03cc effort.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03bf\u03b9\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b7 \u03c0\u03c1\u03ce\u03c4\u03b7 \u03bc\u03ad\u03c4\u03c1\u03b7\u03c3\u03b7 \u03c0\u03bf\u03c5 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03bc\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03af\u03c1\u03b7\u03c3\u03b7;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0388\u03bd\u03b1 \u03b1\u03bd\u03c4\u03b9\u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03b5\u03c5\u03c4\u03b9\u03ba\u03cc validation set \u03ba\u03b1\u03b9 \u03c4\u03b1\u03c5\u03c4\u03cc\u03c7\u03c1\u03bf\u03bd\u03b7 \u03ba\u03b1\u03c4\u03b1\u03b3\u03c1\u03b1\u03c6\u03ae \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2, generated tokens, latency \u03ba\u03b1\u03b9 \u03ba\u03cc\u03c3\u03c4\u03bf\u03c5\u03c2 \u03b3\u03b9\u03b1 baseline \u03ba\u03b1\u03b9 reasoning variants. \u03a7\u03c9\u03c1\u03af\u03c2 \u03b1\u03c5\u03c4\u03cc \u03c4\u03bf \u03b6\u03b5\u03c5\u03b3\u03ac\u03c1\u03b9 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 \u03ba\u03b1\u03b9 \u03ba\u03b1\u03c4\u03b1\u03bd\u03ac\u03bb\u03c9\u03c3\u03b7\u03c2, \u03b7 \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03b2\u03b1\u03c3\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03b5\u03bd\u03c4\u03cd\u03c0\u03c9\u03c3\u03b7.<\/p>\n<\/div>\n<\/details>\n<\/div>\n<\/section>\n<div class=\"td-source-list\">\n<p id=\"piges\" class=\"td-source-list-title\">Izvori<\/p>\n<ul>\n<li><a href=\"https:\/\/arxiv.org\/abs\/2608.26235\" target=\"_blank\" rel=\"noopener noreferrer\">Wani, Dholakia \u03ba\u03b1\u03b9 Ellison, The Reasoning Tax: Token Economics of LLM Reasoning Across Task Types and Deployment Contexts, arXiv:2608.26235<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/Sachin-Wani\/reasoning_tax\" target=\"_blank\" rel=\"noopener noreferrer\">\u0395\u03c0\u03af\u03c3\u03b7\u03bc\u03bf repository \u03c4\u03b7\u03c2 \u03ad\u03c1\u03b5\u03c5\u03bd\u03b1\u03c2 \u03bc\u03b5 dataset, \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1 \u03b1\u03bd\u03ac\u03bb\u03c5\u03c3\u03b7\u03c2, \u03c0\u03af\u03bd\u03b1\u03ba\u03b5\u03c2 \u03ba\u03b1\u03b9 figures<\/a><\/li>\n<li><a href=\"https:\/\/lenovopress.lenovo.com\/lp2368-on-premise-vs-cloud-generative-ai-total-cost-of-ownership-2026-edition\" target=\"_blank\" rel=\"noopener noreferrer\">Lenovo Press, On-Premise vs Cloud: Generative AI Total Cost of Ownership, 2026 Edition<\/a><\/li>\n<li><a href=\"https:\/\/artificialanalysis.ai\/\" target=\"_blank\" rel=\"noopener noreferrer\">Artificial Analysis, benchmark scores, token counts \u03ba\u03b1\u03b9 pricing data \u03c0\u03bf\u03c5 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03ae\u03b8\u03b7\u03ba\u03b1\u03bd \u03c3\u03c4\u03b7 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7<\/a><\/li>\n<\/ul>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>\u03a4\u03bf reasoning tax \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03c0\u03cc\u03c4\u03b5 \u03c4\u03b1 \u03c0\u03c1\u03cc\u03c3\u03b8\u03b5\u03c4\u03b1 reasoning tokens \u03b2\u03b5\u03bb\u03c4\u03b9\u03ce\u03bd\u03bf\u03c5\u03bd \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03b9\u03bc\u03b1 \u03c4\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03ba\u03b1\u03b9 \u03c0\u03cc\u03c4\u03b5 \u03b1\u03c5\u03be\u03ac\u03bd\u03bf\u03c5\u03bd \u03bc\u03cc\u03bd\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2, latency \u03ba\u03b1\u03b9 \u03c5\u03c0\u03bf\u03b4\u03bf\u03bc\u03ae.<\/p>","protected":false},"author":1,"featured_media":113978,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_gspb_post_css":"","footnotes":""},"categories":[199],"tags":[6506,21222,21863,21070,21864],"class_list":["post-105388","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-techniti-noimosyni","tag-ai-agents","tag-ai-ypodomi","tag-llm-kostos","tag-reasoning-models","tag-token-economy-score"],"_links":{"self":[{"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/posts\/105388","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/comments?post=105388"}],"version-history":[{"count":0,"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/posts\/105388\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/media\/113978"}],"wp:attachment":[{"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/media?parent=105388"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/categories?post=105388"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/twodots.gr\/hr\/wp-json\/wp\/v2\/tags?post=105388"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}