Agentic benchmark
Terminal-Bench 3.0 leaderboard
Every model the catalog carries a published Terminal-Bench 3.0 value for, ranked by that value.
CategoryAgentic
MeasureTask completion rate
Tasks74 professional computer-work tasks across 7 domains
DifficultyFrontier autonomous knowledge work
Published byTerminal-Bench 3.0
A continuously maintained benchmark for difficult computer work, including coding, deep learning, finance, engineering, math, and science tasks.
Terminal-Bench 3.0 ranking
11 models with a published Terminal-Bench 3.0 value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Task completion rate |
|---|---|---|---|
| 1 |  Claude Opus 5 | Anthropic | 42.7 |
| 2 |  GPT-5.6 Sol | OpenAI | 34.6 |
| 3 | Claude Fable 5 | Anthropic | 34.0 |
| 4 |  Grok 4.6 | xAI | 26.5 |
| 5 | Claude Opus 4.8 | Anthropic | 21.1 |
| 6 | GPT-5.6 Terra | OpenAI | 20.8 |
| 7 | Grok 4.5 | xAI | 15.7 |
| 8 |  Gemini 3.7 Flash | 14.9 | |
| 9 | Claude Sonnet 5 | Anthropic | 14.6 |
| 10 | GPT-5.6 Luna | OpenAI | 14.3 |
| 11 |  GLM-5.2 | Z.AI | 4.6 |
Evidence key: Observed
Rows are ordered by the value Terminal-Bench 3.0 published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.