New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
Terminal-Bench 2.1 benchmark
AI model leaderboard for the Terminal-Bench 2.1 benchmark. Compare how large language models score on Terminal-Bench 2.1, see the full ranking, and understand what this AI benchmark measures. Claude Fable 5 currently leads with 84.3%. Revised Terminal-Bench 2 suite of real terminal/agentic engineering tasks.
Leaderboard
| # | Model | Organization | Score | Variant | Source |
|---|---|---|---|---|---|
| #1 | Claude Fable 5 | Anthropic | 84.3% | official ↗ | |
| #2 | Claude Mythos 5 | Anthropic | 84.3% | shared with Claude Fable 5 (same underlying model) | official ↗ |
| #3 | Gemini 3.5 Flash | Google DeepMind | 76.2% | official ↗ | |
| #4 | Step 3.7 Flash | StepFun | 59.55% | official ↗ | |
| #5 | NVIDIA Nemotron 3 Ultra | NVIDIA | 56.4% | official ↗ |
Frequently asked questions about Terminal-Bench 2.1
What is the Terminal-Bench 2.1 benchmark?
Revised Terminal-Bench 2 suite of real terminal/agentic engineering tasks.
How is the Terminal-Bench 2.1 benchmark scored?
Terminal-Bench 2.1 is scored using the accuracy (%) metric, where a higher score is better. The maximum achievable score is 100.000. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.
Which AI model scores highest on Terminal-Bench 2.1?
As of the latest reported scores on GenAIList, Claude Fable 5 achieves the highest result on Terminal-Bench 2.1 with a score of 84.3%.
Is a higher Terminal-Bench 2.1 score better?
Yes. On Terminal-Bench 2.1 a higher score indicates better performance, so models near the top of the leaderboard are the strongest.