Skip to content

Terminal-Bench 3.0

A continuously maintained benchmark for difficult computer work, including coding, deep learning, finance, engineering, math, and science tasks.

13 models

Higher is better.

  1. 1Claude Opus 5Anthropic42.7
  2. 2GPT-5.6 SolOpenAI34.6
  3. 3Claude Fable 5Anthropic34.0
  4. 4Grok 4.6xAI26.5
  5. 5Claude Opus 4.8Anthropic21.1
  6. 6GPT-5.6 TerraOpenAI20.8
  7. 7Grok 4.5xAI15.7
  8. 8Gemini 3.7 FlashGoogle14.9
  9. 9Claude Sonnet 5Anthropic14.6
  10. 10GPT-5.6 LunaOpenAI14.3
  11. 11Ornith-1.5-397BOrnith AI13.5
  12. 12Ornith-1.5-35B-A3BOrnith AI5.1
  13. 13GLM-5.2Z.AI4.6
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Task completion rate
Tasks
74 professional computer-work tasks across 7 domains
Authors
Ryan Marten, Alex Shaw, Andy Konwinski, Harbor, and the Laude Institute (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.