Benchmark · Agent tasks
Terminal-Bench
An agent harness running a model in a sandboxed shell, five trials or more on every task, until the task's tests pass. A row is one harness and model pair at one reasoning effort; the model is what's ranked.
- Score
- % of trials passed, higher is better
- Run by
- Stanford and the Laude Institute (the Terminal-Bench maintainers)
- Status
- Current, newest result 20 November 2026
- Updated
- as models are released
- Data
- the published file, Apache-2.0, read 20 November 2026
- Method
- www.tbench.ai/
Results: 4.0
26 models we list, each at its best setting, best first.
| # | Model | Score | Range | Who ran it | Setting | Date | |
|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 A | 64.8% | 61.7% to 68.0% | Independent | effort max, harness Claude Code, harness version 2.1.284, trials 330 | 2026-09-30 | source |
| 2 | Claude Sonnet 5.5 A | 61.8% | 58.9% to 64.8% | Independent | effort max, harness Claude Code, harness version 2.1.284, trials 330 | 2026-09-30 | source |
| 3 | Kimi K3.1 Pro D | 60.9% | 56.8% to 65.0% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-10-30 | source |
| 4 | GPT-6.1 Sol AA | 58.2% | 55.1% to 61.3% | Independent | effort max, harness Codex, harness version 0.159.0, trials 330 | 2026-09-30 | source |
| 5 | Claude Fable 5.1 A | 57.9% | 54.1% to 61.6% | Independent | effort max, harness Claude Code, harness version 2.1.257, trials 330 | 2026-09-02 | source |
| 6 | Gemini 4 Argon B | 54.9% | 51.9% to 57.8% | Independent | effort high, harness Antigravity CLI, harness version 1.5.2, trials 330 | 2026-11-13 | source |
| 7 | GPT-6.1 Terra AA | 51.8% | 49.0% to 54.6% | Independent | effort max, harness Codex, harness version 0.164.0, trials 330 | 2026-10-28 | source |
| 8 | GPT-6 Sol A | 49.4% | 46.2% to 52.6% | Independent | effort max, harness Codex, harness version 0.159.0, trials 330 | 2026-09-30 | source |
| 9 | Muse Wonder D | 47.0% | 43.2% to 50.8% | Independent | effort xhigh, harness Muse Code, harness version 1.4.14, trials 330 | 2026-11-05 | source |
| 10 | GLM-5.3 B | 41.8% | 38.6% to 45.0% | Independent | effort max, harness Claude Code, harness version 2.1.207, trials 330 | 2026-08-26 | source |
| 11 | Claude Haiku 5.5 A | 40.0% | 36.3% to 43.7% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-06 | source |
| 11 | DeepSeek V4.1 Flash C | 40.0% | 36.3% to 43.7% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-10-31 | source |
| 13 | Tencent Hy4 B | 39.1% | 35.6% to 42.6% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-10 | source |
| 14 | GLM-5.3-Flash B | 36.7% | 33.0% to 40.3% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-02 | source |
| 15 | Mistral Large 4 C | 27.9% | 24.0% to 31.8% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-04 | source |
| 16 | Kimi K3 B | 22.4% | 18.6% to 26.2% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-10 | source |
| 17 | Gemini 3.8 Flash AA | 19.1% | 15.7% to 22.4% | Independent | effort high, harness mini-SWE-agent, harness version 2.4.6, trials 330 | 2026-09-01 | source |
| 18 | GPT-6 Luna A | 16.4% | 13.6% to 19.1% | Independent | effort max, harness Codex, harness version 0.159.0, trials 330 | 2026-09-30 | source |
| 19 | Muse Spark 1.3 D | 14.6% | 11.6% to 17.5% | Independent | effort xhigh, harness Muse Code, harness version 1.4.1, trials 330 | 2026-09-30 | source |
| 20 | MiniMax M3 C | 7.3% | 4.6% to 10.0% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-10 | source |
| 21 | DeepSeek V4 Pro D | 6.7% | 4.2% to 9.2% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-10 | source |
| 22 | Beam 1.0 C | 6.4% | 4.7% to 8.0% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-09 | source |
| 23 | Gemini 3.1 Pro (preview) AA | 2.7% | 1.1% to 4.4% | Independent | effort high, harness Antigravity CLI, harness version 1.5.2, trials 330 | 2026-11-14 | source |
| 24 | Inkling-Small B | 0.9% | 0.0% to 1.9% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-06 | source |
| 25 | Nemotron 3.5 Super VL C | 0.3% | -0.3% to 0.9% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-09 | source |
| 26 | Inkling B | 0.0% | 0.0% to 0.0% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-06 | source |
Every setting, 27 rows
| Model | As the leaderboard names it | Score | Who ran it | Setting | Date | |
|---|---|---|---|---|---|---|
| Claude Opus 5.5 | anthropic/claude-opus-5-5 | 64.8% | Independent | effort max, harness Claude Code, harness version 2.1.284, trials 330 | 2026-09-30 | source |
| Claude Sonnet 5.5 | anthropic/claude-sonnet-5-5 | 61.8% | Independent | effort max, harness Claude Code, harness version 2.1.284, trials 330 | 2026-09-30 | source |
| Kimi K3.1 Pro | Kimi K3.1 Pro | 60.9% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-10-30 | source |
| GPT-6.1 Sol | GPT-6.1 Sol | 58.2% | Independent | effort max, harness Codex, harness version 0.159.0, trials 330 | 2026-09-30 | source |
| Claude Fable 5.1 | anthropic/claude-fable-5-1 | 57.9% | Independent | effort max, harness Claude Code, harness version 2.1.257, trials 330 | 2026-09-02 | source |
| Gemini 4 Argon | Gemini 4 Argon | 54.9% | Independent | effort high, harness Antigravity CLI, harness version 1.5.2, trials 330 | 2026-11-13 | source |
| GPT-6.1 Terra | GPT-6.1 Terra | 51.8% | Independent | effort max, harness Codex, harness version 0.164.0, trials 330 | 2026-10-28 | source |
| GPT-6 Sol | GPT-6 Sol | 49.4% | Independent | effort max, harness Codex, harness version 0.159.0, trials 330 | 2026-09-30 | source |
| Muse Wonder | Muse Wonder | 47.0% | Independent | effort xhigh, harness Muse Code, harness version 1.4.14, trials 330 | 2026-11-05 | source |
| GLM-5.3 | GLM-5.3 | 41.8% | Independent | effort max, harness Claude Code, harness version 2.1.207, trials 330 | 2026-08-26 | source |
| Claude Haiku 5.5 | anthropic/claude-haiku-5-5 | 40.0% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-06 | source |
| DeepSeek V4.1 Flash | DeepSeek V4.1 Flash 1017 | 40.0% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-10-31 | source |
| Tencent Hy4 | Hy4 | 39.1% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-10 | source |
| GLM-5.3-Flash | GLM-5.3-Flash | 36.7% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-02 | source |
| GLM-5.3-Flash | GLM-5.3-Flash | 35.8% | Independent | effort none, harness Claude Code, harness version 2.1.285, trials 330 | 2026-10-06 | source |
| Mistral Large 4 | Mistral Large 4 | 27.9% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-04 | source |
| Kimi K3 | Kimi K3 | 22.4% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-10 | source |
| Gemini 3.8 Flash | Gemini 3.8 Flash | 19.1% | Independent | effort high, harness mini-SWE-agent, harness version 2.4.6, trials 330 | 2026-09-01 | source |
| GPT-6 Luna | GPT-6 Luna | 16.4% | Independent | effort max, harness Codex, harness version 0.159.0, trials 330 | 2026-09-30 | source |
| Muse Spark 1.3 | Muse Spark 1.3 | 14.6% | Independent | effort xhigh, harness Muse Code, harness version 1.4.1, trials 330 | 2026-09-30 | source |
| MiniMax M3 | MiniMax M3 | 7.3% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-10 | source |
| DeepSeek V4 Pro | DeepSeek V4 Pro 0813 | 6.7% | Independent | effort max, harness Claude Code, harness version 2.1.311, trials 330 | 2026-11-10 | source |
| Beam 1.0 | Beam 1.0 | 6.4% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-09 | source |
| Gemini 3.1 Pro (preview) | gemini/gemini-3.1-pro-preview | 2.7% | Independent | effort high, harness Antigravity CLI, harness version 1.5.2, trials 330 | 2026-11-14 | source |
| Inkling-Small | Inkling Small | 0.9% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-06 | source |
| Nemotron 3.5 Super VL | Nemotron 3.5 Super VL | 0.3% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-09 | source |
| Inkling | Inkling | 0.0% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, trials 330 | 2026-11-06 | source |
Results: 4.1
24 models we list, each at its best setting, best first.
| # | Model | Score | Range | Who ran it | Setting | Date | |
|---|---|---|---|---|---|---|---|
| 1 | Claude Fable 5.2 A | 65.8% | 62.9% to 68.6% | Independent | effort max, harness Claude Code, harness version 2.1.321, trials 330 | 2026-11-10 | source |
| 2 | Gemini 4 Carbon B | 64.5% | 61.3% to 67.8% | Independent | effort high, harness Antigravity CLI, harness version 1.5.7, trials 330 | 2026-11-14 | source |
| 3 | GPT-6.2 Astra AA | 62.1% | 59.8% to 64.4% | Independent | effort max, harness Codex, harness version 0.169.0, trials 330 | 2026-11-17 | source |
| 4 | GPT-6.2 Sol AA | 61.5% | 58.8% to 64.3% | Independent | effort max, harness Codex, harness version 0.169.0, trials 330 | 2026-11-17 | source |
| 5 | Gemini 4 Neon B | 61.2% | 58.4% to 64.1% | Independent | effort xhigh, harness Antigravity CLI, harness version 1.5.12, trials 330 | 2026-11-20 | source |
| 6 | Kimi K3.1 Pro D | 59.7% | 55.5% to 63.9% | Independent | effort max, harness Claude Code, harness version 2.1.311, regrade yes, trials 330 | 2026-11-12 | source |
| 7 | Gemini 4 Hydrogen B | 56.4% | 53.4% to 59.3% | Independent | effort high, harness Antigravity CLI, harness version 1.5.7, trials 330 | 2026-11-13 | source |
| 8 | Gemini 4 Argon B | 55.5% | 52.6% to 58.3% | Independent | effort high, harness Antigravity CLI, harness version 1.5.2, regrade yes, trials 330 | 2026-11-13 | source |
| 9 | GPT-6.1 Terra AA | 53.3% | 50.2% to 56.4% | Independent | effort max, harness Codex, harness version 0.164.0, regrade yes, trials 330 | 2026-11-12 | source |
| 10 | Muse Wonder D | 47.0% | 43.2% to 50.8% | Independent | effort xhigh, harness Muse Code, harness version 1.4.14, regrade yes, trials 330 | 2026-11-13 | source |
| 11 | Claude Haiku 5.5 A | 40.0% | 36.4% to 43.6% | Independent | effort max, harness Claude Code, harness version 2.1.311, regrade yes, trials 330 | 2026-11-12 | source |
| 12 | DeepSeek V4.1 Flash C | 38.5% | 34.7% to 42.2% | Independent | effort max, harness Claude Code, harness version 2.1.311, regrade yes, trials 330 | 2026-11-12 | source |
| 13 | Tencent Hy4 B | 37.9% | 34.4% to 41.4% | Independent | effort max, harness Claude Code, harness version 2.1.311, regrade yes, trials 330 | 2026-11-12 | source |
| 14 | GLM-5.3-Flash B | 37.0% | 33.3% to 40.6% | Independent | effort max, harness Claude Code, harness version 2.1.311, regrade yes, trials 330 | 2026-11-12 | source |
| 15 | Muse Ember D | 30.6% | 26.8% to 34.4% | Independent | effort xhigh, harness Muse Code, harness version 1.4.14, trials 330 | 2026-11-10 | source |
| 16 | Mistral Large 4 C | 27.6% | 23.7% to 31.4% | Independent | effort max, harness Claude Code, harness version 2.1.311, regrade yes, trials 330 | 2026-11-12 | source |
| 17 | Kimi K3 B | 23.0% | 19.3% to 26.8% | Independent | effort max, harness Claude Code, harness version 2.1.311, regrade yes, trials 330 | 2026-11-12 | source |
| 18 | MiniMax M3 C | 7.6% | 4.8% to 10.3% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, regrade yes, trials 330 | 2026-11-12 | source |
| 19 | DeepSeek V4 Pro D | 6.7% | 4.2% to 9.2% | Independent | effort max, harness Claude Code, harness version 2.1.311, regrade yes, trials 330 | 2026-11-12 | source |
| 20 | Beam 1.0 C | 6.4% | 4.7% to 8.0% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, regrade yes, trials 330 | 2026-11-12 | source |
| 21 | Gemini 3.1 Pro (preview) AA | 2.7% | 1.1% to 4.4% | Independent | effort high, harness Antigravity CLI, harness version 1.5.2, regrade yes, trials 330 | 2026-11-15 | source |
| 22 | Inkling-Small B | 0.9% | 0.0% to 1.9% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, regrade yes, trials 330 | 2026-11-12 | source |
| 23 | Nemotron 3.5 Super VL C | 0.3% | -0.3% to 0.9% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, regrade yes, trials 330 | 2026-11-12 | source |
| 24 | Inkling B | 0.0% | 0.0% to 0.0% | Independent | effort none, harness mini-swe-agent, harness version 2.4.6, regrade yes, trials 330 | 2026-11-12 | source |
A rank is among the models we list with a result, not a place on the whole leaderboard, and a model offered by several hosts is shown once, on its own listing. These results don't move our grades. Data: Terminal-Bench, Stanford and the Laude Institute, Apache 2.0, Apache-2.0.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.