Terminal-Bench 4.1
We show this table for reference; we do not rank on it.
The current Terminal-Bench release keeps the 66 professional computer-work tasks of 4.0 and regrades every saved trial with verifiers hardened against reward hacking and overly strict checks.
All releases
This page carries every Terminal-Bench release: 4.1 above, then the 4.0, 3.0, 2.1 and 2.0 tables and a cross-version ranking. Versions that change the task set or resources are not a like-for-like model comparison.
Jump to version historyTasks completed on Terminal-Bench 4.1 — November 10, 2026 snapshot
We mirror the published tasks completed view for Terminal-Bench 4.1. Claude Opus 5.5 leads the public snapshot at 65.76%, followed by Claude Sonnet 5.5 (63.94%) and GPT-6.1 Sol (60.91%). We do not use these results to rank models overall.
1Closed
Claude Opus 5.5
Anthropic
Claude Code · max reasoning
65.76%
Overall 85.77Context 1M
2Closed
Claude Sonnet 5.5
Anthropic
Claude Code · max reasoning
63.94%
Overall 81.24Context 1M
3Closed
GPT-6.1 Sol
OpenAI
Codex · max reasoning
60.91%
Overall 84.91Context 1.05M
35 modelsAgenticCurrentDisplay onlyUpdated November 10, 2026 snapshot
Current releaseTerminal-Bench 4.1 · Released · 66 tasks · Official leaderboard · Regrades the 4.0 trials
Tasks completed table (35 models)
35 of 35 rows
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 5.5AnthropicClaude Code · max reasoning | 65.76% | Not reported | Closed |
| 2 | Claude Sonnet 5.5AnthropicClaude Code · max reasoning | 63.94% | Not reported | Closed |
| 3 | GPT-6.1 SolOpenAICodex · max reasoning | 60.91% | Not reported | Closed |
| 4 | GPT-6 AstraOpenAICodex · max reasoning | 60.30% | Not reported | Closed |
| 5 | GPT-6 AstraOpenAICodex · high reasoning | 60.00% | Not reported | Closed |
| 6 | GPT-6 AstraOpenAICodex · xhigh reasoning | 59.70% | Not reported | Closed |
| 7 | Claude Fable 5.1AnthropicClaude Code · xhigh reasoning | 57.58% | Not reported | Closed |
| 8 | Claude Fable 5.1AnthropicClaude Code · max reasoning | 57.27% | Not reported | Closed |
| 9 | GPT-6 AstraOpenAICodex · medium reasoning | 55.15% | Not reported | Closed |
| 10 | Claude Fable 5.1AnthropicClaude Code · high reasoning | 54.85% | Not reported | Closed |
| 11 | Claude Fable 5.1AnthropicClaude Code · medium reasoning | 53.94% | Not reported | Closed |
| 12 | Claude Opus 5AnthropicClaude Code · xhigh reasoning | 53.33% | Not reported | Closed |
| 13 | Claude Opus 5AnthropicClaude Code · max reasoning | 52.12% | Not reported | Closed |
| 14 | GPT-6 AstraOpenAICodex · low reasoning | 50.91% | Not reported | Closed |
| 15 | GPT-6 SolOpenAICodex · max reasoning | 50.30% | Not reported | Closed |
| 16 | Claude Opus 5AnthropicClaude Code · high reasoning | 49.70% | Not reported | Closed |
| 17 | Claude Opus 5AnthropicClaude Code · medium reasoning | 45.15% | Not reported | Closed |
| 18 | Claude Fable 5AnthropicClaude Code · max reasoning | 44.85% | Not reported | Closed |
| 19 | Claude Fable 5.1AnthropicClaude Code · low reasoning | 43.33% | Not reported | Closed |
| 20 | GLM-5.3Z.AIClaude Code · max reasoning | 41.21% | Not reported | Open |
| 21 | Grok 4.7xAIGrok Build · xhigh reasoning | 38.79% | Not reported | Closed |
| 22 | GPT-5.6 SolOpenAICodex · max reasoning | 38.48% | Not reported | Closed |
| 23 | GLM-5.3-FlashZ.AIClaude Code · none reasoning | 35.45% | Not reported | Open |
| 24 | Claude Opus 5AnthropicClaude Code · low reasoning | 34.85% | Not reported | Closed |
| 25 | Qwen3.8-Max-0902AlibabaClaude Code · max reasoning | 26.97% | Not reported | Unknown |
| 26 | Claude Opus 4.8AnthropicClaude Code · max reasoning | 23.64% | Not reported | Closed |
| 27 | GPT-5.6 TerraOpenAICodex · max reasoning | 21.52% | Not reported | Closed |
| 28 | Grok 4.6xAIGrok Build · high reasoning | 19.70% | Not reported | Closed |
| 29 | Gemini 3.8 FlashGooglemini-SWE-agent · high reasoning | 18.18% | Not reported | Closed |
| 30 | GPT-5.6 LunaOpenAICodex · max reasoning | 17.58% | Not reported | Closed |
| 31 | GPT-6 LunaOpenAICodex · max reasoning | 15.15% | Not reported | Closed |
| 32 | Muse Spark 1.3MetaMuse Code · xhigh reasoning | 14.85% | Not reported | Closed |
| 33 | Gemini 3.7 FlashGooglemini-SWE-agent · high reasoning | 12.42% | Not reported | Closed |
| 34 | Grok 4.5xAIGrok Build · high reasoning | 12.42% | Not reported | Closed |
| 35 | Claude Sonnet 5AnthropicClaude Code · max reasoning | 12.12% | Not reported | Closed |
Previous releaseReleased · 66 tasks · Official leaderboard · Not comparable with 3.0 or earlier
Terminal-Bench 4.0
We mirror the official 4.0 table. The raw results compare complete model-and-agent systems. The latest normalized scores replace 3.0 within one external agentic benchmark family and cannot independently make a category ranking-eligible.
35 of 35 rows
| Source rank | Source model or configuration | Tasks completed | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 64.85% | Claude Code · max reasoning · Original result | Not reported | Closed |
| 2 | Claude Sonnet 5.5Anthropic | 61.82% | Claude Code · max reasoning · Original result | Not reported | Closed |
| 3 | GPT-6 AstraOpenAI | 58.18% | Codex · max reasoning · Original result | Not reported | Closed |
| 3 | GPT-6.1 SolOpenAI | 58.18% | Codex · max reasoning · Original result | Not reported | Closed |
| 5 | Claude Fable 5.1Anthropic | 57.88% | Claude Code · max reasoning · Original result | Not reported | Closed |
| 5 | GPT-6 AstraOpenAI | 57.88% | Codex · xhigh reasoning · Original result | Not reported | Closed |
| 5 | GPT-6 AstraOpenAI | 57.88% | Codex · high reasoning · Original result | Not reported | Closed |
| 5 | Claude Fable 5.1Anthropic | 57.88% | Claude Code · xhigh reasoning · Original result | Not reported | Closed |
| 9 | Claude Fable 5.1Anthropic | 54.55% | Claude Code · high reasoning · Original result | Not reported | Closed |
| 10 | GPT-6 AstraOpenAI | 54.24% | Codex · medium reasoning · Original result | Not reported | Closed |
| 11 | Claude Fable 5.1Anthropic | 53.94% | Claude Code · medium reasoning · Original result | Not reported | Closed |
| 11 | Claude Opus 5Anthropic | 53.94% | Claude Code · xhigh reasoning · Original result | Not reported | Closed |
| 13 | Claude Opus 5Anthropic | 51.82% | Claude Code · max reasoning · Original result | Not reported | Closed |
| 14 | GPT-6 AstraOpenAI | 50.61% | Codex · low reasoning · Original result | Not reported | Closed |
| 15 | Claude Opus 5Anthropic | 50.30% | Claude Code · high reasoning · Original result | Not reported | Closed |
| 16 | GPT-6 SolOpenAI | 49.39% | Codex · max reasoning · Original result | Not reported | Closed |
| 17 | Claude Opus 5Anthropic | 44.85% | Claude Code · medium reasoning · Original result | Not reported | Closed |
| 18 | Claude Fable 5Anthropic | 44.55% | Claude Code · max reasoning · Original result | Not reported | Closed |
| 19 | Claude Fable 5.1Anthropic | 43.33% | Claude Code · low reasoning · Original result | Not reported | Closed |
| 20 | GLM-5.3Z.AI | 41.82% | Claude Code · max reasoning · Original result | Not reported | Open |
| 21 | Grok 4.7xAI | 37.58% | Grok Build · xhigh reasoning · Original result | Not reported | Closed |
| 22 | GPT-5.6 SolOpenAI | 37.27% | Codex · max reasoning · Original result | Not reported | Closed |
| 23 | GLM-5.3-FlashZ.AI | 35.76% | Claude Code · none reasoning · Original result | Not reported | Open |
| 24 | Claude Opus 5Anthropic | 34.85% | Claude Code · low reasoning · Original result | Not reported | Closed |
| 25 | Qwen3.8-Max-0902Alibaba | 26.97% | Claude Code · max reasoning · Original result | Not reported | Unknown |
| 26 | Claude Opus 4.8Anthropic | 23.64% | Claude Code · max reasoning · Original result | Not reported | Closed |
| 27 | GPT-5.6 TerraOpenAI | 21.52% | Codex · max reasoning · Original result | Not reported | Closed |
| 28 | Grok 4.6xAI | 20.30% | Grok Build · high reasoning · Original result | Not reported | Closed |
| 29 | Gemini 3.8 FlashGoogle | 19.09% | mini-SWE-agent · high reasoning · Original result | Not reported | Closed |
| 30 | GPT-5.6 LunaOpenAI | 17.27% | Codex · max reasoning · Original result | Not reported | Closed |
| 31 | GPT-6 LunaOpenAI | 16.36% | Codex · max reasoning · Original result | Not reported | Closed |
| 32 | Muse Spark 1.3Meta | 14.55% | Muse Code · xhigh reasoning · Original result | Not reported | Closed |
| 33 | Grok 4.5xAI | 12.42% | Grok Build · high reasoning · Original result | Not reported | Closed |
| 33 | Claude Sonnet 5Anthropic | 12.42% | Claude Code · max reasoning · Original result | Not reported | Closed |
| 35 | Gemini 3.7 FlashGoogle | 11.21% | mini-SWE-agent · high reasoning · Original result | Not reported | Closed |
AA Terminal-Bench 4.0
Stored separately from the benchmark-owned Terminal-Bench 4.0 lane because Artificial Analysis controls the agent harness and configuration. Display-only; not yet admitted as independent-run evidence.
22 of 22 rows
| Source rank | Source model or configuration | Score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Kimi K3.1 ProMoonshot AI | 67.2% | Artificial Analysis · Original result | Not reported | Open |
| 2 | Gemini 4 CarbonGoogle | 65.2% | Artificial Analysis · Original result | Not reported | Closed |
| 3 | Claude Sonnet 5.5Anthropic | 63.6% | Artificial Analysis · Original result | Not reported | Closed |
| 4 | Claude Fable 5.2Anthropic | 59.6% | Artificial Analysis · Original result | Not reported | Closed |
| 5 | Claude Opus 5.5Anthropic | 59.6% | Artificial Analysis · Original result | Not reported | Closed |
| 6 | GPT-6 AstraOpenAI | 59.1% | Artificial Analysis · Original result | Not reported | Closed |
| 7 | Muse Wonder FullMeta | 57.1% | Artificial Analysis · Original result | Not reported | Closed |
| 8 | GPT-6.1 SolOpenAI | 56.1% | Artificial Analysis · Original result | Not reported | Closed |
| 9 | GPT-6.1 TerraOpenAI | 55.6% | Artificial Analysis · Original result | Not reported | Closed |
| 10 | Claude Fable 5.1Anthropic | 52.0% | Artificial Analysis · Original result | Not reported | Closed |
| 11 | Muse WonderMeta | 52.0% | Artificial Analysis · Original result | Not reported | Closed |
| 12 | GLM-5.3Z.AI | 41.9% | Artificial Analysis · Original result | Not reported | Open |
| 13 | Muse EmberMeta | 40.9% | Artificial Analysis · Original result | Not reported | Closed |
| 14 | MiMo-V2.6-ProXiaomi | 34.8% | Artificial Analysis · Original result | Not reported | Open |
| 15 | Claude Haiku 5.5Anthropic | 32.8% | Artificial Analysis · Original result | Not reported | Closed |
| 16 | Grok 4.7xAI | 25.8% | Artificial Analysis · Original result | Not reported | Closed |
| 17 | Mistral Large 4Mistral | 24.7% | Artificial Analysis · Original result | Not reported | Open |
| 18 | Gemini 3.8 FlashGoogle | 19.7% | Artificial Analysis · Original result | Not reported | Closed |
| 19 | Inkling 2 SmallThinking Machines Lab | 13.1% | Artificial Analysis · Original result | Not reported | Closed |
| 20 | GPT-6 LunaOpenAI | 12.6% | Artificial Analysis · Original result | Not reported | Closed |
| 21 | Qwen3.8-27BAlibaba | 5.6% | Artificial Analysis · Original result | Not reported | Open |
| 22 | Beam 1.0 501B-A23BReflection AI | 3.0% | Artificial Analysis · Original result | Not reported | Open |
Terminal-Bench 4.0 (Vals)
We keep the Vals-hosted Terminal-Bench 4.0 table on a dedicated key so it never overwrites the official Terminal-Bench 4.0 leaderboard rows. Vals runs its own harness and publishes per-domain splits; this mirror is display only and does not enter weighted rankings.
63 of 63 rows
| Source rank | Source model or configuration | Vals Terminal-Bench 4.0 mirror score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Fable 5.2Anthropic | 65.15% | Vals AI · Original result | Not reported | Closed |
| 2 | Claude Opus 5.5Anthropic | 65.15% | Vals AI · Original result | Not reported | Closed |
| 3 | Claude Sonnet 5.5Anthropic | 64.14% | Vals AI · Original result | Not reported | Closed |
| 4 | Kimi K3.1 ProMoonshot AI | 63.64% | max reasoning · Original result | Not reported | Open |
| 5 | Gemini 4 CarbonGoogle | 63.13% | high reasoning · Original result | Not reported | Closed |
| 6 | GPT-6 AstraOpenAI | 59.60% | max reasoning · Original result | Not reported | Closed |
| 7 | Claude Fable 5.1Anthropic | 58.08% | Vals AI · Original result | Not reported | Closed |
| 8 | Muse Wonder FullMeta | 58.08% | max reasoning · Original result | Not reported | Closed |
| 9 | Gemini 4 ArgonGoogle | 57.58% | high reasoning · Original result | Not reported | Closed |
| 10 | GPT-6.1 SolOpenAI | 55.05% | max reasoning · Original result | Not reported | Closed |
| 11 | Claude Opus 5Anthropic | 53.53% | Vals AI · Original result | Not reported | Closed |
| 12 | GPT-6.1 TerraOpenAI | 49.49% | max reasoning · Original result | Not reported | Closed |
| 13 | Muse WonderMeta | 49.49% | max reasoning · Original result | Not reported | Closed |
| 14 | GPT-6 SolOpenAI | 44.44% | max reasoning · Original result | Not reported | Closed |
| 15 | Claude Fable 5Anthropic | 41.41% | Vals AI · Original result | Not reported | Closed |
| 16 | Muse EmberMeta | 39.39% | max reasoning · Original result | Not reported | Closed |
| 17 | Hy4Tencent | 39.39% | Vals AI · Original result | Not reported | Pending |
| 18 | GLM-5.3Z.AI | 38.89% | max reasoning · Original result | Not reported | Open |
| 19 | Qwen3.8 Max InternalAlibaba | 38.89% | Vals AI · Original result | Not reported | Unknown |
| 20 | GPT-5.6 SolOpenAI | 37.88% | max reasoning · Original result | Not reported | Closed |
| 21 | GLM-5.3-FlashZ.AI | 37.37% | max reasoning · Original result | Not reported | Open |
| 22 | Claude Haiku 5.5Anthropic | 36.36% | Vals AI · Original result | Not reported | Closed |
| 23 | Qwen3.8 MaxAlibaba | 34.34% | Vals AI · Original result | Not reported | Open |
| 24 | DeepSeek V4.1 Flash 1017DeepSeek | 31.82% | max reasoning · Original result | Not reported | Open |
| 25 | Step 5 PreviewStepFun | 31.82% | Vals AI · Original result | Not reported | Open |
| 26 | MiMo-V2.6-ProXiaomi | 31.31% | Vals AI · Original result | Not reported | Open |
| 27 | Grok 4.7xAI | 28.79% | xhigh reasoning · Original result | Not reported | Closed |
| 28 | Mistral Large 4Mistral | 27.78% | high reasoning · Original result | Not reported | Open |
| 29 | Mistral Large 4 OssMistral AI | 26.26% | high reasoning · Original result | Not reported | Unknown |
| 30 | MiniMax M3.1 FlashMiniMax | 25.25% | Vals AI · Original result | Not reported | Closed |
| 31 | Muse Spark 1.3 MaxMeta | 24.75% | max reasoning · Original result | Not reported | Unknown |
| 32 | MiMo-V2.6-FlashXiaomi | 24.24% | Vals AI · Original result | Not reported | Open |
| 33 | Claude Opus 4.8Anthropic | 23.23% | Vals AI · Original result | Not reported | Closed |
| 34 | GPT-5.6 TerraOpenAI | 22.73% | max reasoning · Original result | Not reported | Closed |
| 35 | Ember-1Fireworks | 20.71% | Vals AI · Original result | Not reported | Closed |
| 36 | DeepSeek V4.1 Flash 1017DeepSeek | 19.70% | high reasoning · Original result | Not reported | Open |
| 37 | Gemini 3.8 FlashGoogle | 19.70% | high reasoning · Original result | Not reported | Closed |
| 38 | DeepSeek V4 Flash 0731DeepSeek | 18.69% | high reasoning · Original result | Not reported | Open |
| 39 | Grok 4.6xAI | 17.17% | high reasoning · Original result | Not reported | Closed |
| 40 | Kimi K3Moonshot AI | 17.17% | max reasoning · Original result | Not reported | Open |
| 41 | DeepSeek V4 Pro 0813DeepSeek | 14.14% | max reasoning · Original result | Not reported | Open |
| 42 | GPT-6 LunaOpenAI | 13.64% | max reasoning · Original result | Not reported | Closed |
| 43 | Inkling 2 SmallThinking Machines Lab | 12.63% | 0.99 reasoning · Original result | Not reported | Closed |
| 44 | Gemini 3.7 FlashGoogle | 12.12% | high reasoning · Original result | Not reported | Closed |
| 45 | GPT-5.6 LunaOpenAI | 11.62% | max reasoning · Original result | Not reported | Closed |
| 46 | DeepSeek V4 Pro 0813DeepSeek | 11.11% | max reasoning · Original result | Not reported | Open |
| 47 | Claude Sonnet 5Anthropic | 10.61% | Vals AI · Original result | Not reported | Closed |
| 48 | Muse Spark 1.3Meta | 10.61% | xhigh reasoning · Original result | Not reported | Closed |
| 49 | Grok 4.5xAI | 8.59% | high reasoning · Original result | Not reported | Closed |
| 50 | Hy4 previewTencent | 8.08% | Vals AI · Original result | Not reported | Open |
| 51 | Muse Spark OpenMeta | 6.57% | xhigh reasoning · Original result | Not reported | Open |
| 52 | Gemini 3.5 FlashGoogle | 6.06% | high reasoning · Original result | Not reported | Closed |
| 53 | Muse Spark 1.2Meta | 6.06% | xhigh reasoning · Original result | Not reported | Closed |
| 54 | Beam 1.0Reflection | 4.04% | Vals AI · Original result | Not reported | Unknown |
| 55 | Mercury 3Inception | 3.03% | high reasoning · Original result | Not reported | Closed |
| 56 | Gemini 3.1 Pro PreviewGoogle | 2.52% | high reasoning · Original result | Not reported | Unknown |
| 57 | GPT-5.4 miniOpenAI | 2.52% | xhigh reasoning · Original result | Not reported | Closed |
| 58 | MiniMax M3MiniMax | 1.01% | Vals AI · Original result | Not reported | Open |
| 59 | InklingThinking Machines Lab | 0.51% | 0.99 reasoning · Original result | Not reported | Open |
| 60 | MiMo-V2.5-ProXiaomi | 0.51% | Vals AI · Original result | Not reported | Closed |
| 61 | Inkling-SmallThinking Machines Lab | 0.00% | 0.99 reasoning · Original result | Not reported | Open |
| 62 | Nemotron 3.5 Super VL 120B A12BNVIDIA | 0.00% | high reasoning · Original result | Not reported | Open |
| 63 | Mercury 2.5Inception | 0.00% | high reasoning · Original result | Not reported | Closed |
Terminal-Bench version history
Every Terminal-Bench release, newest first. Scores are not directly comparable across releases that change the task set or resources (2.x, 3.0 and 4.0). Version 4.1 is a regrade of the 4.0 trials, so its scores stay within each entry’s 4.0 confidence interval.
| Version | Released | Tasks | What changed | Leader | Models |
|---|---|---|---|---|---|
| Terminal-Bench 4.1Current | Nov 9, 2026 | 66 tasks | Regrades the saved 4.0 trials with verifiers hardened against reward hacking; no agent-facing changes. | LeaderClaude Opus 5.5 65.8%Official board, Claude Code | 23 models |
| Terminal-Bench 4.0 | Aug 28, 2026 | 66 tasks | Recalibrates task resources, fixes 19 tasks and removes 8; fresh trials, not comparable with 3.0. | LeaderClaude Opus 5.5 64.8%Official board, Claude Code | 23 models |
| Terminal-Bench 3.0 | Jul 30, 2026 | 74 tasks, 7 domains | New professional computer-work task set, first published as FrontierBench. | LeaderClaude Opus 5 42.7%Official board, mini-SWE-agent | 12 models |
| Terminal-Bench 2.1 | May 6, 2026 | 89 tasks | Fixes 28 Terminal-Bench 2.0 tasks. No official leaderboard; results are provider reports. | LeaderSWE-2 92.8%Highest provider report | 87 models |
| Terminal-Bench 2.0 | Nov 7, 2025 | 89 tasks | Harder terminal task set released with the Harbor evaluation framework. | LeaderGPT-5.5 84.7%Official board, NexAU-AHE | 48 models |
Rankings across Terminal-Bench versions
Each cell is the model’s rank among models on that release’s official leaderboard, using its best published agent and effort setting, with the score below. Rows cover every model on the 4.1, 4.0 or 3.0 boards. Version 2.1 has no official leaderboard, so it has no column here. A dash means the model has no row on that board.
| Model | Terminal-Bench 4.123 models | Terminal-Bench 4.023 models | Terminal-Bench 3.012 models |
|---|---|---|---|
| Claude Opus 5.5 | #165.8% | #164.8% | Not on this board |
| Claude Sonnet 5.5 | #263.9% | #261.8% | Not on this board |
| GPT-6.1 Sol | #360.9% | #358.2% | Not on this board |
| GPT-6 Astra | 1 (Down 1 place from 4.0)#460.3% | #358.2% | Not on this board |
| Claude Fable 5.1 | #557.6% | #557.9% | Not on this board |
| Claude Opus 5 | #653.3% | #653.9% | #142.7% |
| GPT-6 Sol | #750.3% | #749.4% | Not on this board |
| Claude Fable 5 | #844.9% | #844.5% | #334.0% |
| GLM-5.3 | #941.2% | #941.8% | #432.4% |
| Grok 4.7 | #1038.8% | #1037.6% | Not on this board |
| GPT-5.6 Sol | #1138.5% | #1137.3% | #234.6% |
| GLM-5.3-Flash | #1235.5% | #1235.8% | Not on this board |
| Qwen3.8-Max-0902 | #1327.0% | #1327.0% | Not on this board |
| Claude Opus 4.8 | #1423.6% | #1423.6% | #621.1% |
| GPT-5.6 Terra | #1521.5% | #1521.5% | #720.8% |
| Grok 4.6 | #1619.7% | #1620.3% | #526.5% |
| Gemini 3.8 Flash | #1718.2% | #1719.1% | Not on this board |
| GPT-5.6 Luna | #1817.6% | #1817.3% | #1114.3% |
| GPT-6 Luna | #1915.2% | #1916.4% | Not on this board |
| Muse Spark 1.3 | #2014.8% | #2014.6% | Not on this board |
| Grok 4.5 | #2112.4% | #2112.4% | #915.7% |
| Gemini 3.7 Flash | 2 (Up 2 places from 4.0)#2112.4% | #2311.2% | Not on this board |
| Claude Sonnet 5 | 2 (Down 2 places from 4.0)#2312.1% | #2112.4% | #1014.6% |
| SWE-1.7 Lightning | Not on this board | Not on this board | #818.6% |
| GLM-5.2 | Not on this board | Not on this board | #124.6% |
Places gained or lost from 4.0 to 4.1. Version 4.1 regrades the 4.0 trials, so this is the one like-for-like rank change. 3 of 23 models moved.
No model above has a row on the 2.0 board (48 models), so 2.0 has no column. See the Terminal-Bench 2.0 table.
Older releaseReleased · 74 tasks, 7 domains · Official leaderboard · Not comparable with other releases
Terminal-Bench 3.0
We mirror the official version 0.1 table. The raw results compare complete model-and-agent systems. Their normalized scores count as one external agentic benchmark family and cannot independently make a category ranking-eligible.
12 of 12 rows
| Source rank | Source model or configuration | Tasks completed | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Opus 5Anthropic | 42.7% | max reasoning · Original result | Not reported | Closed |
| 2 | GPT-5.6 SolOpenAI | 34.6% | max reasoning · Original result | Not reported | Closed |
| 3 | Claude Fable 5Anthropic | 34.0% | max reasoning · Original result | Not reported | Closed |
| 4 | GLM-5.3Z.AI | 32.4% | max reasoning · Original result | Not reported | Open |
| 5 | Grok 4.6xAI | 26.5% | high reasoning · Original result | Not reported | Closed |
| 6 | Claude Opus 4.8Anthropic | 21.1% | max reasoning · Original result | Not reported | Closed |
| 7 | GPT-5.6 TerraOpenAI | 20.8% | max reasoning · Original result | Not reported | Closed |
| 8 | SWE-1.7 LightningCognition | 18.6% | Ryan Marten, Alex Shaw, Andy Konwinski, Harbor, and the Laude Institute · Original result | Not reported | Unknown |
| 9 | Grok 4.5xAI | 15.7% | xhigh reasoning · Original result | Not reported | Closed |
| 10 | Claude Sonnet 5Anthropic | 14.6% | max reasoning · Original result | Not reported | Closed |
| 11 | GPT-5.6 LunaOpenAI | 14.3% | max reasoning · Original result | Not reported | Closed |
| 12 | GLM-5.2Z.AI | 4.6% | max reasoning · Original result | Not reported | Open |
Older releaseReleased · 89 tasks · Provider reports · Not comparable with other releases
Terminal-Bench 2.1
BenchLM stores Terminal-Bench 2.1 results on their own key, apart from Terminal-Bench 2.0, because the two versions are not directly comparable. Harness and effort settings vary by row.
87 of 87 rows
| Source rank | Source model or configuration | Score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | SWE-2Cognition | 92.8% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 2 | GPT-5.6 SolOpenAI | 91.9% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 3 | DeepSeek V4.1 Flash 1017DeepSeek | 90.6% | Terminal-Bench contributors · Original result | Not reported | Open |
| 4 | MiMo-V2.6-ProXiaomi | 89.9% | Terminal-Bench contributors · Original result | Not reported | Open |
| 5 | Gemini 3.8 FlashGoogle | 89.4% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 6 | SWE-2 FlashCognition | 89.2% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 7 | Muse Spark 1.3Meta | 88.8% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 8 | LongCat-2.5Meituan | 88.4% | Terminal-Bench contributors · Original result | Not reported | Open |
| 9 | Kimi K3Moonshot AI | 88.3% | Terminal-Bench contributors · Original result | Not reported | Open |
| 10 | GLM-5.3Z.AI | 88.2% | Terminal-Bench contributors · Original result | Not reported | Open |
| 11 | Claude Mythos 5Anthropic | 88.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 12 | DeepSeek V4 Pro 0813DeepSeek | 87.9% | Terminal-Bench contributors · Original result | Not reported | Open |
| 13 | MiMo-V2.6-FlashXiaomi | 87.6% | Terminal-Bench contributors · Original result | Not reported | Open |
| 14 | GPT-5.6 TerraOpenAI | 87.4% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 15 | Qwen3.8 MaxAlibaba | 86.6% | Terminal-Bench contributors · Original result | Not reported | Open |
| 16 | Ornith-1.5-397BOrnith AI | 86.1% | Terminal-Bench contributors · Original result | Not reported | Open |
| 17 | Gemini 3.7 FlashGoogle | 85.8% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 18 | Hy4 previewTencent | 85.4% | Terminal-Bench contributors · Original result | Not reported | Open |
| 19 | Inkling 2 SmallThinking Machines Lab | 85.1% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 20 | Step 5 PreviewStepFun | 85.0% | Terminal-Bench contributors · Original result | Not reported | Open |
| 21 | GPT-5.6 LunaOpenAI | 84.7% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 22 | Claude Fable 5Anthropic | 84.3% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 23 | GLM-5.3-FlashZ.AI | 84.3% | Terminal-Bench contributors · Original result | Not reported | Open |
| 24 | Grok 4.5xAI | 83.3% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 25 | Ultramarine 1 FlashPaper Instruments | 83.2% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 26 | Muse Spark 1.2Meta | 82.9% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 27 | DeepSeek V4 Flash 0731DeepSeek | 82.7% | Terminal-Bench contributors · Original result | Not reported | Open |
| 28 | Sakana Fugu-UltraSakana AI | 82.1% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 29 | Ember-1Fireworks | 82.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 30 | SWE-1.7Cognition | 81.5% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 31 | Ling 3.1 FlashInclusionAI | 81.2% | Terminal-Bench contributors · Original result | Not reported | Open |
| 32 | GLM-5.2Z.AI | 81.0% | Terminal-Bench contributors · Original result | Not reported | Open |
| 33 | Ripple-65B PreviewRicursive Intelligence | 80.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 34 | Claude Sonnet 5Anthropic | 80.4% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 35 | Sakana FuguSakana AI | 80.2% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 36 | Beam 1.0 501B-A23BReflection AI | 80.1% | Terminal-Bench contributors · Original result | Not reported | Open |
| 37 | Muse Spark 1.1Meta | 80.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 38 | Particle-1Particle AI | 79.9% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 39 | Beam 1.0 501B-A23B FP8Reflection AI | 78.9% | Terminal-Bench contributors · Original result | Not reported | Open |
| 40 | Beam 1.0 501B-A23B NVFP4Reflection AI | 78.4% | Terminal-Bench contributors · Original result | Not reported | Open |
| 41 | Atria Dawn PreviewShanghai Artificial Intelligence Laboratory | 78.3% | Terminal-Bench contributors · Original result | Not reported | Open |
| 42 | Ornith-1.0-397BDeepReinforce AI | 77.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 43 | Gemini 3.5 FlashGoogle | 76.2% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 44 | dots3-note PreviewDots Studio | 75.1% | Terminal-Bench contributors · Original result | Not reported | Open |
| 45 | Claude Opus 4.8Anthropic | 74.6% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 46 | Jais-3-375B-A23BIFM | 73.8% | Terminal-Bench contributors · Original result | Not reported | Open |
| 47 | Qwen3.8-27BAlibaba | 73.0% | Terminal-Bench contributors · Original result | Not reported | Open |
| 48 | OrcaCyber-Nano-1.0OrcaRouter | 72.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 49 | Seed 2.1 ProByteDance | 71.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 50 | Apodex 1.1Apodex | 70.8% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 51 | Laguna S 2.1Poolside | 70.2% | Terminal-Bench contributors · Original result | Not reported | Open |
| 52 | Quasar 438BMultiverse Computing | 69.3% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 53 | Ornith-1.5-35B-A3BOrnith AI | 67.8% | Terminal-Bench contributors · Original result | Not reported | Open |
| 54 | Seed 2.1 TurboByteDance | 67.6% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 55 | Jais-3-MoVA-36B-A4BIFM | 67.3% | Terminal-Bench contributors · Original result | Not reported | Open |
| 56 | Nemotron 3.5 Super VL 120B A12BNVIDIA | 66.7% | Terminal-Bench contributors · Original result | Not reported | Open |
| 57 | MiniMax M3MiniMax | 66.0% | Terminal-Bench contributors · Original result | Not reported | Open |
| 58 | Pokee-Isaac 28BPokee AI | 65.1% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 59 | Inkling-SmallThinking Machines Lab | 64.7% | Terminal-Bench contributors · Original result | Not reported | Open |
| 60 | Ornith-1.0-35BDeepReinforce AI | 64.2% | Terminal-Bench contributors · Original result | Not reported | Open |
| 61 | InklingThinking Machines Lab | 63.8% | Terminal-Bench contributors · Original result | Not reported | Open |
| 62 | Jais-3-32BIFM | 63.7% | Terminal-Bench contributors · Original result | Not reported | Open |
| 63 | MAI-Code-1.1-FlashMicrosoft | 62.9% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 64 | Nemotron 3.5 Super VL 120B A12B NVFP4NVIDIA | 60.8% | Terminal-Bench contributors · Original result | Not reported | Open |
| 65 | Step 3.7 FlashStepFun | 59.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 66 | Ling 3.0 FlashInclusionAI | 57.0% | Terminal-Bench contributors · Original result | Not reported | Open |
| 67 | Solar Pro 4Upstage | 57.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 68 | Nemotron 3 UltraNVIDIA | 56.4% | Terminal-Bench contributors · Original result | Not reported | Open |
| 69 | FlintRecursive | 54.9% | Terminal-Bench contributors · Original result | Not reported | Open |
| 70 | Gemini 3.5 Flash-LiteGoogle | 54.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 71 | Nemotron 3.5 Bolt 74B-A9B BF16NVIDIA | 53.9% | Terminal-Bench contributors · Original result | Not reported | Open |
| 72 | Ternary Bonsai 2 27BPrism ML | 52.8% | Terminal-Bench contributors · Original result | Not reported | Open |
| 73 | Muse Glimmer 30BMeta | 51.7% | Terminal-Bench contributors · Original result | Not reported | Open |
| 74 | Nemotron 3.5 Bolt 74B-A9BNVIDIA | 51.0% | Terminal-Bench contributors · Original result | Not reported | Open |
| 75 | Jais-3-7BIFM | 46.4% | Terminal-Bench contributors · Original result | Not reported | Open |
| 76 | Ornith-1.5-9BOrnith AI | 46.2% | Terminal-Bench contributors · Original result | Not reported | Open |
| 77 | K-EXAONE 2.0LG AI Research | 43.8% | Terminal-Bench contributors · Original result | Not reported | Open |
| 78 | Ornith-1.0-9BDeepReinforce AI | 43.1% | Terminal-Bench contributors · Original result | Not reported | Open |
| 79 | Jais-3-3.7BIFM | 41.2% | Terminal-Bench contributors · Original result | Not reported | Open |
| 80 | A.X K2SK Telecom | 36.0% | Terminal-Bench contributors · Original result | Not reported | Open |
| 81 | Granite 4.2 30BIBM | 29.2% | Terminal-Bench contributors · Original result | Not reported | Open |
| 82 | Mercury 2Inception | 27.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 83 | Nemotron 3.5 Lightning 30B A3B NVFP4NVIDIA | 23.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 84 | Granite 4.2 8BIBM | 20.6% | Terminal-Bench contributors · Original result | Not reported | Open |
| 85 | Mellum2.1-12B-A2.5B-ThinkingJetBrains | 17.4% | Terminal-Bench contributors · Original result | Not reported | Open |
| 86 | LFM3-1.7BLiquidAI | 12.9% | Terminal-Bench contributors · Original result | Not reported | Open |
| 87 | MiniCPM5-2BOpenBMB | 8.6% | Terminal-Bench contributors · Original result | Not reported | Open |
AA Terminal-Bench 2.1
Stored separately from the core Terminal-Bench lane because Artificial Analysis controls the evaluation configuration.
21 of 21 rows
| Source rank | Source model or configuration | Score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Fable 5.1Anthropic | 91.4% | Artificial Analysis · Original result | Not reported | Closed |
| 2 | Claude Opus 5Anthropic | 89.1% | Artificial Analysis · Original result | Not reported | Closed |
| 3 | GPT-6 AstraOpenAI | 88.4% | Artificial Analysis · Original result | Not reported | Closed |
| 4 | Grok 4.6xAI | 88.4% | Artificial Analysis · Original result | Not reported | Closed |
| 5 | GPT-5.6 TerraOpenAI | 88.0% | Artificial Analysis · Original result | Not reported | Closed |
| 6 | Gemini 3.8 FlashGoogle | 87.6% | Artificial Analysis · Original result | Not reported | Closed |
| 7 | Qwen3.8-Flash-NextAlibaba | 86.1% | Artificial Analysis · Original result | Not reported | Open |
| 8 | Gemini 3.7 FlashGoogle | 85.8% | Artificial Analysis · Original result | Not reported | Closed |
| 9 | Kimi K3Moonshot AI | 85.0% | Artificial Analysis · Original result | Not reported | Open |
| 10 | Claude Fable 5Anthropic | 84.6% | Artificial Analysis · Original result | Not reported | Closed |
| 11 | Claude Opus 4.8Anthropic | 84.6% | Artificial Analysis · Original result | Not reported | Closed |
| 12 | GPT-5.5OpenAI | 84.3% | Artificial Analysis · Original result | Not reported | Closed |
| 13 | GLM-5.3-FlashZ.AI | 84.3% | Artificial Analysis · Original result | Not reported | Open |
| 14 | GLM-5.3Z.AI | 83.9% | Artificial Analysis · Original result | Not reported | Open |
| 15 | Claude Opus 4.7 (Adaptive)Anthropic | 83.1% | Artificial Analysis · Original result | Not reported | Closed |
| 16 | Grok 4.5xAI | 81.6% | Artificial Analysis · Original result | Not reported | Closed |
| 17 | Inkling 2 SmallThinking Machines Lab | 81.3% | Artificial Analysis · Original result | Not reported | Closed |
| 18 | GPT-5.6 LunaOpenAI | 80.9% | Artificial Analysis · Original result | Not reported | Closed |
| 19 | Claude Sonnet 5Anthropic | 80.5% | Artificial Analysis · Original result | Not reported | Closed |
| 20 | Muse Spark 1.2Meta | 80.1% | Artificial Analysis · Original result | Not reported | Closed |
| 21 | Qwen3.8-27BAlibaba | 79.8% | Artificial Analysis · Original result | Not reported | Open |
Terminal-Bench 2.1 (Vals)
BenchLM mirrors this Vals AI leaderboard for reference only. It is excluded from BenchLM overall and category rankings.
76 of 76 rows
| Source rank | Source model or configuration | Vals Terminal-Bench 2.1 mirror score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 87.6% | Vals AI · Original result | Not reported | Closed |
| 2 | GPT-6 AstraOpenAI | 87.3% | max reasoning · Original result | Not reported | Closed |
| 3 | GPT-5.6 SolOpenAI | 85.8% | max reasoning · Original result | Not reported | Closed |
| 4 | Claude Fable 5.1Anthropic | 85.0% | Vals AI · Original result | Not reported | Closed |
| 5 | Claude Opus 5Anthropic | 84.6% | Vals AI · Original result | Not reported | Closed |
| 6 | GPT-6 SolOpenAI | 83.1% | max reasoning · Original result | Not reported | Closed |
| 7 | Claude Sonnet 5.5Anthropic | 83.1% | Vals AI · Original result | Not reported | Closed |
| 8 | Gemini 3.8 FlashGoogle | 81.3% | high reasoning · Original result | Not reported | Closed |
| 9 | Kimi K3Moonshot AI | 80.9% | Vals AI · Original result | Not reported | Open |
| 10 | Claude Fable 5Anthropic | 80.5% | Vals AI · Original result | Not reported | Closed |
| 11 | GPT-5.6 LunaOpenAI | 79.0% | max reasoning · Original result | Not reported | Closed |
| 12 | Muse Spark 1.3 MaxMeta | 79.0% | max reasoning · Original result | Not reported | Unknown |
| 13 | Grok 4.6xAI | 78.3% | high reasoning · Original result | Not reported | Closed |
| 14 | Gemini 3.7 FlashGoogle | 77.5% | high reasoning · Original result | Not reported | Closed |
| 15 | GPT-5.6 TerraOpenAI | 77.5% | max reasoning · Original result | Not reported | Closed |
| 16 | GPT-5.5OpenAI | 76.4% | high reasoning · Original result | Not reported | Closed |
| 17 | MiMo-V2.6-FlashXiaomi | 76.4% | Vals AI · Original result | Not reported | Open |
| 18 | DeepSeek V4.1 Flash 1017DeepSeek | 74.5% | high reasoning · Original result | Not reported | Open |
| 19 | Claude Sonnet 5Anthropic | 74.5% | Vals AI · Original result | Not reported | Closed |
| 20 | Gemini 3.5 FlashGoogle | 74.2% | high reasoning · Original result | Not reported | Closed |
| 21 | Gemini 3.6 FlashGoogle | 73.8% | high reasoning · Original result | Not reported | Closed |
| 22 | Grok 4.7xAI | 73.4% | xhigh reasoning · Original result | Not reported | Closed |
| 23 | GPT-6 LunaOpenAI | 73.0% | max reasoning · Original result | Not reported | Closed |
| 24 | Muse Spark 1.3Meta | 72.3% | xhigh reasoning · Original result | Not reported | Closed |
| 25 | Claude Opus 4.8Anthropic | 71.9% | Vals AI · Original result | Not reported | Closed |
| 26 | GLM-5.3Z.AI | 71.5% | max reasoning · Original result | Not reported | Open |
| 27 | Gemini 3.1 Pro PreviewGoogle | 70.8% | high reasoning · Original result | Not reported | Unknown |
| 28 | Claude Opus 4.8 Claude CodeAnthropic | 69.7% | Claude Code · Original result | Not reported | Unknown |
| 29 | Muse Spark 1.2Meta | 69.7% | xhigh reasoning · Original result | Not reported | Closed |
| 30 | Muse Spark 1.1Meta | 69.3% | xhigh reasoning · Original result | Not reported | Closed |
| 31 | Claude Opus 4.7Anthropic | 68.5% | Vals AI · Original result | Not reported | Closed |
| 32 | Grok 4.5xAI | 67.8% | high reasoning · Original result | Not reported | Closed |
| 33 | GLM-5.2Z.AI | 67.8% | max reasoning · Original result | Not reported | Open |
| 34 | MiMo-V2.6-ProXiaomi | 67.8% | Vals AI · Original result | Not reported | Open |
| 35 | Qwen3.8 MaxAlibaba | 67.4% | Vals AI · Original result | Not reported | Open |
| 36 | DeepSeek V4 Flash 0731DeepSeek | 67.0% | Vals AI · Original result | Not reported | Open |
| 37 | Kimi K2.7 CodeMoonshot AI | 67.0% | Vals AI · Original result | Not reported | Open |
| 38 | GLM-5.3-FlashZ.AI | 62.9% | max reasoning · Original result | Not reported | Open |
| 39 | Qwen3.7 MaxAlibaba | 61.0% | Vals AI · Original result | Not reported | Closed |
| 40 | MiMo-V2.5Xiaomi | 60.7% | Vals AI · Original result | Not reported | Closed |
| 41 | Composer 2.5Cursor | 58.4% | Cursor CLI · Original result | Not reported | Closed |
| 42 | Qwen3.8-27BAlibaba | 58.4% | xhigh reasoning · Original result | Not reported | Open |
| 43 | Gemini 4 ArgonGoogle | 57.7% | high reasoning · Original result | Not reported | Closed |
| 44 | GPT-5.5 CodexOpenAI | 57.3% | Codex · Original result | Not reported | Unknown |
| 45 | GPT-5.5 FactoryOpenAI | 57.3% | Factory · Original result | Not reported | Unknown |
| 46 | Claude Sonnet 4.6Anthropic | 57.3% | Vals AI · Original result | Not reported | Closed |
| 47 | MiMo-V2.5-ProXiaomi | 57.3% | Vals AI · Original result | Not reported | Closed |
| 48 | GLM-5.1Z.AI | 56.9% | Vals AI · Original result | Not reported | Open |
| 49 | Inkling-SmallThinking Machines Lab | 55.1% | 0.99 reasoning · Original result | Not reported | Open |
| 50 | Hy4 previewTencent | 55.1% | Vals AI · Original result | Not reported | Open |
| 51 | GPT-5.4 miniOpenAI | 54.7% | high reasoning · Original result | Not reported | Closed |
| 52 | DeepSeek V4 Pro 0813DeepSeek | 54.7% | max reasoning · Original result | Not reported | Open |
| 53 | Gemini 3 Flash PreviewGoogle | 53.9% | high reasoning · Original result | Not reported | Unknown |
| 54 | Kimi K2.6Moonshot AI | 53.6% | Vals AI · Original result | Not reported | Open |
| 55 | MiniMax M3MiniMax | 53.6% | Vals AI · Original result | Not reported | Open |
| 56 | Qwen3.6 PlusAlibaba | 53.2% | Vals AI · Original result | Not reported | Closed |
| 57 | Qwen3.7 PlusAlibaba | 52.8% | Vals AI · Original result | Not reported | Closed |
| 58 | Nemotron 3 Ultra 550b A55bNvidia | 50.9% | Vals AI · Original result | Not reported | Unknown |
| 59 | Gemini 3.5 Flash-LiteGoogle | 50.2% | high reasoning · Original result | Not reported | Closed |
| 60 | Ling 3.0 Flash 2607Ant | 50.2% | Vals AI · Original result | Not reported | Unknown |
| 61 | Ling 3.0 Flash Af Rc3Ant | 50.2% | Vals AI · Original result | Not reported | Unknown |
| 62 | DeepSeek V4 Pro 0813DeepSeek | 50.2% | Vals AI · Original result | Not reported | Open |
| 63 | MiniMax M2.7MiniMax | 48.7% | Vals AI · Original result | Not reported | Open |
| 64 | InklingThinking Machines Lab | 47.6% | 0.99 reasoning · Original result | Not reported | Open |
| 65 | Grok 4.20 0309 ReasoningxAI | 44.2% | Vals AI · Original result | Not reported | Unknown |
| 66 | Claude Haiku 4.5 ThinkingAnthropic | 43.8% | Vals AI · Original result | Not reported | Closed |
| 67 | Grok 4.3xAI | 41.9% | high reasoning · Original result | Not reported | Closed |
| 68 | Kimi K2.5 ThinkingMoonshot AI | 41.9% | Vals AI · Original result | Not reported | Unknown |
| 69 | GPT-5.4 nanoOpenAI | 41.6% | high reasoning · Original result | Not reported | Closed |
| 70 | Mistral Medium 3.5Mistral AI | 39.0% | high reasoning · Original result | Not reported | Unknown |
| 71 | Mercury 2.5Inception | 34.5% | high reasoning · Original result | Not reported | Closed |
| 72 | Gemini 3.1 Flash Lite PreviewGoogle | 34.1% | high reasoning · Original result | Not reported | Unknown |
| 73 | Laguna M.1Poolside | 34.1% | Vals AI · Original result | Not reported | Closed |
| 74 | Laguna XS.2Poolside | 25.8% | Vals AI · Original result | Not reported | Open |
| 75 | Command A Plus 05 2026Cohere | 17.6% | Vals AI · Original result | Not reported | Unknown |
| 76 | Nemotron Lightning 3p5 30b A3bFireworks AI | 10.9% | Vals AI · Original result | Not reported | Unknown |
Older releaseReleased · 89 tasks · Official leaderboard · Not comparable with other releases
Terminal-Bench 2.0
Terminal-Bench 2.0 focuses on realistic CLI and repository workflows rather than toy code generation. BenchLM scores it from the agentic rows, which the benchmark page shows; this coding entry lists only the models whose coding record carries the result.
45 of 45 rows
| Source rank | Source model or configuration | Score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | GPT-5.5OpenAI | 82.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 2 | GPT-5.3 CodexOpenAI | 77.3% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 3 | GPT-5.4OpenAI | 75.1% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 4 | Qwen3.7 PlusAlibaba | 70.3% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 5 | Qwen3.7 MaxAlibaba | 69.7% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 6 | Claude Opus 4.7 (Adaptive)Anthropic | 69.4% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 7 | Composer 2.5Cursor | 69.3% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 8 | MiMo-V2.5-ProXiaomi | 68.4% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 9 | DeepSeek V4 Pro 0813DeepSeek | 67.9% | Terminal-Bench contributors · Original result | Not reported | Open |
| 10 | Kimi K2.6Moonshot AI | 66.7% | Terminal-Bench contributors · Original result | Not reported | Open |
| 11 | MiMo-V2.5Xiaomi | 65.8% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 12 | Claude Opus 4.6Anthropic | 65.4% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 13 | Qwen 3.6 Max (preview)Alibaba | 65.4% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 14 | GLM-5.1Z.AI | 63.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 15 | DeepSeek V4 Pro (High)DeepSeek | 63.3% | Terminal-Bench contributors · Original result | Not reported | Open |
| 16 | Composer 2Cursor | 61.7% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 17 | Qwen3.6 PlusAlibaba | 61.6% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 18 | GPT-5.4 miniOpenAI | 60.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 19 | Qwen3.6-27BAlibaba | 59.3% | Terminal-Bench contributors · Original result | Not reported | Open |
| 20 | Claude Opus 4.5Anthropic | 59.3% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 21 | Claude Sonnet 4.6Anthropic | 59.1% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 22 | DeepSeek V4 ProDeepSeek | 59.1% | Terminal-Bench contributors · Original result | Not reported | Open |
| 23 | Muse SparkMeta | 59.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 24 | MiniMax M2.7MiniMax | 57.0% | Terminal-Bench contributors · Original result | Not reported | Open |
| 25 | DeepSeek V4 Flash 0731DeepSeek | 56.9% | Terminal-Bench contributors · Original result | Not reported | Open |
| 26 | DeepSeek V4 Flash (High)DeepSeek | 56.6% | Terminal-Bench contributors · Original result | Not reported | Open |
| 27 | GLM-5Z.AI | 56.2% | Terminal-Bench contributors · Original result | Not reported | Open |
| 28 | Hy3 PreviewTencent | 54.4% | Terminal-Bench contributors · Original result | Not reported | Open |
| 29 | Qwen3.5 397BAlibaba | 52.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 30 | Qwen3.6-35B-A3BAlibaba | 51.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 31 | Kimi K2.5 (Reasoning)Moonshot AI | 50.8% | Terminal-Bench contributors · Original result | Not reported | Open |
| 32 | Kimi K2.5Moonshot AI | 50.8% | Terminal-Bench contributors · Original result | Not reported | Open |
| 33 | Claude Sonnet 4.5Anthropic | 50.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 34 | Qwen3.5-122B-A10BAlibaba | 49.4% | Terminal-Bench contributors · Original result | Not reported | Open |
| 35 | DeepSeek V4 FlashDeepSeek | 49.1% | Terminal-Bench contributors · Original result | Not reported | Open |
| 36 | Grok 4.20xAI | 47.1% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 37 | GPT-5.4 nanoOpenAI | 46.3% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 38 | MAI-Thinking-1Microsoft | 46.0% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 39 | Laguna M.1Poolside | 45.8% | Terminal-Bench contributors · Original result | Not reported | Closed |
| 40 | Qwen3.5-27BAlibaba | 41.6% | Terminal-Bench contributors · Original result | Not reported | Open |
| 41 | GLM-4.7Z.AI | 41.0% | Terminal-Bench contributors · Original result | Not reported | Open |
| 42 | Qwen3.5-35B-A3BAlibaba | 40.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 43 | Laguna XS 2.1Poolside | 37.5% | Terminal-Bench contributors · Original result | Not reported | Open |
| 44 | Laguna XS.2Poolside | 35.7% | Terminal-Bench contributors · Original result | Not reported | Open |
| 45 | LongCat-Flash-Lite-SparseMeituan | 33.7% | Terminal-Bench contributors · Original result | Not reported | Open |
Vals Terminal-Bench 2.0 mirror
BenchLM mirrors this Vals AI leaderboard for reference only. It is excluded from BenchLM overall and category rankings.
67 of 67 rows
| Source rank | Source model or configuration | Vals Terminal-Bench 2.0 mirror score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | GPT-5.5OpenAI | 73.20% | high reasoning · Original result | Not reported | Closed |
| 2 | Claude Opus 4.8Anthropic | 70.04% | Vals AI · Original result | Not reported | Closed |
| 3 | Claude Opus 4.7Anthropic | 68.54% | Vals AI · Original result | Not reported | Closed |
| 4 | Gemini 3.5 FlashGoogle | 67.42% | high reasoning · Original result | Not reported | Closed |
| 5 | Gemini 3.1 Pro PreviewGoogle | 67.42% | high reasoning · Original result | Not reported | Unknown |
| 6 | GPT-5.3 CodexOpenAI | 64.05% | high reasoning · Original result | Not reported | Closed |
| 7 | Muse SparkMeta | 59.55% | Vals AI · Original result | Not reported | Closed |
| 8 | Claude Sonnet 4.6Anthropic | 59.55% | Vals AI · Original result | Not reported | Closed |
| 9 | Qwen3.7 MaxAlibaba | 59.18% | Vals AI · Original result | Not reported | Closed |
| 10 | Claude Opus 4.5Anthropic | 58.43% | Vals AI · Original result | Not reported | Closed |
| 11 | Claude Opus 4.6 (Adaptive)Anthropic | 58.43% | Vals AI · Original result | Not reported | Closed |
| 12 | GPT-5.4OpenAI | 58.43% | high reasoning · Original result | Not reported | Closed |
| 13 | Kimi K2.6Moonshot AI | 57.30% | Vals AI · Original result | Not reported | Open |
| 14 | DeepSeek V4 Pro 0813DeepSeek | 56.18% | Vals AI · Original result | Not reported | Open |
| 15 | Gemini 3 Pro PreviewGoogle | 55.06% | high reasoning · Original result | Not reported | Unknown |
| 16 | Claude Opus 4.5 ThinkingAnthropic | 53.93% | Vals AI · Original result | Not reported | Closed |
| 17 | GLM-5.1Z.AI | 53.93% | Vals AI · Original result | Not reported | Open |
| 18 | Gemini 3 Flash PreviewGoogle | 51.69% | high reasoning · Original result | Not reported | Unknown |
| 19 | GPT-5.2OpenAI | 51.69% | high reasoning · Original result | Not reported | Closed |
| 20 | Qwen 3.6 Max (preview)Alibaba | 51.69% | Vals AI · Original result | Not reported | Closed |
| 21 | GLM 5 ThinkingZhipu AI | 49.44% | Vals AI · Original result | Not reported | Unknown |
| 22 | MiniMax M2.7MiniMax | 47.19% | Vals AI · Original result | Not reported | Open |
| 23 | MiniMax M3MiniMax | 46.07% | Vals AI · Original result | Not reported | Open |
| 24 | GPT-5.4 miniOpenAI | 44.94% | high reasoning · Original result | Not reported | Closed |
| 25 | Qwen3.6 PlusAlibaba | 44.94% | Vals AI · Original result | Not reported | Closed |
| 26 | GPT-5.1OpenAI | 44.94% | high reasoning · Original result | Not reported | Closed |
| 27 | Qwen3.6-27BAlibaba | 44.94% | Vals AI · Original result | Not reported | Open |
| 28 | Grok 4.3xAI | 43.45% | Vals AI · Original result | Not reported | Closed |
| 29 | Claude Sonnet 4.5 ThinkingAnthropic | 41.57% | Vals AI · Original result | Not reported | Closed |
| 30 | MiniMax M2.5MiniMax | 41.57% | Vals AI · Original result | Not reported | Closed |
| 31 | Qwen3.5 Plus ThinkingAlibaba | 41.57% | Vals AI · Original result | Not reported | Unknown |
| 32 | Grok 4.20 0309 ReasoningxAI | 40.45% | Vals AI · Original result | Not reported | Unknown |
| 33 | Kimi K2.5 ThinkingMoonshot AI | 40.45% | Vals AI · Original result | Not reported | Unknown |
| 34 | GPT-5.4 nanoOpenAI | 39.89% | high reasoning · Original result | Not reported | Closed |
| 35 | Gemma 4 31b ItGoogle | 39.33% | high reasoning · Original result | Not reported | Unknown |
| 36 | Claude Haiku 4.5 ThinkingAnthropic | 38.20% | Vals AI · Original result | Not reported | Closed |
| 37 | GLM-4.7Z.AI | 38.20% | Vals AI · Original result | Not reported | Open |
| 38 | MiniMax M2.1MiniMax | 37.08% | Vals AI · Original result | Not reported | Unknown |
| 39 | GPT-5OpenAI | 37.08% | high reasoning · Original result | Not reported | Unknown |
| 40 | Kimi K2 ThinkingMoonshot AI | 37.08% | Vals AI · Original result | Not reported | Unknown |
| 41 | DeepSeek V3p2 ThinkingFireworks AI | 35.95% | high reasoning · Original result | Not reported | Unknown |
| 42 | DeepSeek V3p2Fireworks AI | 34.83% | none reasoning · Original result | Not reported | Unknown |
| 43 | Laguna M.1Poolside | 31.46% | Vals AI · Original result | Not reported | Closed |
| 44 | Gemini 2.5 ProGoogle | 30.34% | Vals AI · Original result | Not reported | Closed |
| 45 | Mistral Medium 3.5Mistral AI | 30.34% | high reasoning · Original result | Not reported | Unknown |
| 46 | Grok 4 Fast (Reasoning)xAI | 29.21% | Vals AI · Original result | Not reported | Closed |
| 47 | Laguna XS.2Poolside | 28.09% | Vals AI · Original result | Not reported | Open |
| 48 | GLM-4.6Z.AI | 28.09% | Vals AI · Original result | Not reported | Open |
| 49 | Grok 4 0709xAI | 28.09% | Vals AI · Original result | Not reported | Unknown |
| 50 | GPT-5 miniOpenAI | 26.97% | high reasoning · Original result | Not reported | Closed |
| 51 | Moonshotai Kimi K2 InstructTogether AI | 25.84% | Vals AI · Original result | Not reported | Unknown |
| 52 | Grok 4.1 Fast (Reasoning)xAI | 24.72% | Vals AI · Original result | Not reported | Closed |
| 53 | Gemini 3.1 Flash Lite PreviewGoogle | 24.72% | high reasoning · Original result | Not reported | Unknown |
| 54 | Qwen3.5 FlashAlibaba | 24.72% | Vals AI · Original result | Not reported | Closed |
| 55 | Qwen3 MaxAlibaba | 24.72% | Vals AI · Original result | Not reported | Closed |
| 56 | DeepSeek V3p1Fireworks AI | 22.47% | Vals AI · Original result | Not reported | Unknown |
| 57 | Gemini 2.5 Flash Preview 09 2025 ThinkingGoogle | 21.35% | Vals AI · Original result | Not reported | Unknown |
| 58 | Qwen3 MaxAlibaba | 20.23% | Vals AI · Original result | Not reported | Closed |
| 59 | GPT-OSS 120BOpenAI | 19.10% | Vals AI · Original result | Not reported | Open |
| 60 | Grok 4.1 Fast Non ReasoningxAI | 17.98% | Vals AI · Original result | Not reported | Unknown |
| 61 | Trinity-Large-ThinkingArcee AI | 17.98% | Vals AI · Original result | Not reported | Open |
| 62 | Mistral Small 2603Mistral AI | 16.85% | high reasoning · Original result | Not reported | Unknown |
| 63 | GPT-4.1OpenAI | 14.61% | high reasoning · Original result | Not reported | Closed |
| 64 | Magistral Medium 2509Mistral AI | 13.48% | Vals AI · Original result | Not reported | Unknown |
| 65 | Mistral Large 2512Mistral AI | 8.99% | Vals AI · Original result | Not reported | Unknown |
| 66 | Command A 03 2025Cohere | 2.25% | Vals AI · Original result | Not reported | Unknown |
| 67 | Llama4 Maverick Instruct BasicFireworks AI | 2.25% | Vals AI · Original result | Not reported | Unknown |
How to read this leaderboard
Operator receipt: 35 sourced rows are currently displayable on this page; the leading published row is Claude Opus 5.5 at 65.76%.
Honest limit: The table compares complete model-and-agent systems, not isolated base models. Errored trials, trials without stored artifacts, and 28 regrades attributed to verifier measurement noise keep their 4.0 scores, so 4.1 is a regrade of 4.0 rather than a fresh run.
How we use Terminal-Bench 4.1
Terminal-Bench 4.1 keeps the 66 tasks, 5 trials per task and 8-hour timeout of 4.0 and regrades the saved trials with hardened verifiers. Agent code now runs as an untrusted user in a separate process, and several over-strict checks accept valid alternative solutions. 162 of 11,550 trials changed score, and every entry stayed within its 4.0 95% confidence interval.
Each row measures a model and agent harness together, such as Claude Code, Codex, or Grok Build. We mirror the table display-only. The ranking still reads the 4.0 table for the Terminal-Bench family, so the regrade does not move any BenchLM score.
Snapshot
35 system rows66 tasks5 trials per taskRegrade of 4.0Display only
The published Terminal-Bench 4.1 snapshot places Claude Opus 5.5 first at 65.76%. The third row is 4.85 points behind. The broader top-10 range is 10.91 points, so the table still separates the published systems.
35 models have been evaluated on Terminal-Bench 4.1. The benchmark falls in the Agentic category. Terminal-Bench 4.1 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Terminal-Bench 4.1
Year
2026
Tasks
66 professional computer-work tasks
Format
Task completion rate across 5 trials per task
Difficulty
Frontier autonomous knowledge work
Version 4.1 makes no agent-facing changes. It re-scores the saved 4.0 artifacts with updated verifiers: agent code now runs as an untrusted user in a separate process, and several over-strict checks accept valid alternative solutions. 162 of 11,550 trials changed score, and every one of the 35 leaderboard entries stayed within its 4.0 95% confidence interval. BenchLM keeps 4.1 on its own key so the 4.0 table remains a fixed record.
Freshness and provenance
Version
Terminal-Bench 4.1.0
Refresh cadence
Continuous
Staleness state
Current
Question availability
Public benchmark set
CurrentDisplay only
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does Terminal-Bench 4.1 measure?
The current Terminal-Bench release keeps the 66 professional computer-work tasks of 4.0 and regrades every saved trial with verifiers hardened against reward hacking and overly strict checks.
Which model leads the published Terminal-Bench 4.1 snapshot?
Claude Opus 5.5 currently leads the published Terminal-Bench 4.1 snapshot with 65.76% tasks completed. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on Terminal-Bench 4.1?
The November 10, 2026 snapshot snapshot contains 35 AI models.
Last updated: November 10, 2026 snapshot · mirrored from the public benchmark leaderboard
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.