Luciano Ayres

Enterprise AI in Numbers · Focus area

AI model benchmarks, prices and choices

Which model leads which job, at what price, and from how many credible vendors?

Three labs, Anthropic, OpenAI and Google DeepMind, are statistically tied at the top of Epoch AI's capabilities index, and the lead changes by job: GPT-6 Astra leads coding on FrontierSWE at 65.5%, while Gemini 4 Argon leads professional services work at 82.4%. Capability that was the best available a year ago now lists at $0.07 per million tokens, and the best Chinese model trails the best US model by 11 points.

15 indicators · Updated

Which model leads which job

The best independent result on each benchmark so far, and the model and lab that hold it.

JobLeading modelScoreSinceEvaluator
Coding benchmark leaderGPT-6 Astra OpenAI65.5%September 3, 2026Epoch AI, Benchmarking hub
Professional work leaderGemini 4 Argon Google DeepMind82.4%September 30, 2026Epoch AI, Benchmarking hub
Remote work leaderClaude Opus 5.5 Anthropic21.3%September 22, 2026Epoch AI, Benchmarking hub
Computer use leaderClaude Opus 5 Anthropic31.4%July 24, 2026Epoch AI, Benchmarking hub
Expert reasoning leaderGPT-6 Astra OpenAI54.8%September 3, 2026Epoch AI, Benchmarking hub
Factual accuracy leaderGemini 4 Argon Google DeepMind76.0%September 30, 2026Epoch AI, Benchmarking hub
Most preferred modelGemini 4 Argon (High) Google DeepMind1,525.7 pointsOctober 22, 2026LMArena, leaderboard dataset

More indicators

Dates ahead

  1. Claude Sonnet 4.5 is retired from the Claude API · Anthropic
  2. OpenAI shuts down GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 pro, o3 and o3-pro in the API · OpenAI
  3. OpenAI shuts down GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini in the API · OpenAI
  4. Google's earliest shutdown date for Gemini 2.5 Flash Image in the Gemini API · Google
  5. OpenAI shuts down GPT-5.1, GPT-5.3-Codex and GPT-5.4 nano in the API · OpenAI
  6. Google's earliest shutdown date for Gemini 3.1 Flash-Lite in the Gemini API · Google

More from the data