We just open sourced a tiny GPT-style cognitive core built in pure Rust.See our repository
Terminal-Bench 2.1
One harness to unlock the potential of all models
Compare Ante runs across models on the same Terminal-Bench 2.1 task set, using consistent parameters and verified benchmark results.
Best Accuracy85.8%
Leading Model
Grok 4.5
Grok 4.5Task Set89 tasks
Trials380 passed / 445 trials
Benchmark principlesRead more
We benchmark what we ship.Every eval uses a pinned public . No eval-only branches or benchmark-specific prompts.
The runs are auditable.Every result links its raw Harbor run, so anyone can inspect the trials behind the number.
The constraints are official.All trials follow the official Terminal-Bench parameters: 89 tasks, 5 trials per task, strict timeouts, and hardware limits.
Model org
All model orgsTB 2.1 · Same parameters: 89 tasks · 5 trials/task · Updated Jul 10, 2026
| # | Model | Same-model | Agent | Source | |||||
|---|---|---|---|---|---|---|---|---|---|
| 1 | Grok 4.5 | 85.8% ±1.66 SE | $242.57 | 8.4 min | #1 same-model | Ante0.preview.56 | Harbor link | Jul 10, 2026 | |
| 2 | GLM 5.2 | 74.6% ±2.06 SE | $260.11 | 11.3 min | #1 same-model | Ante0.preview.43 | Harbor link | Jun 20, 2026 | |
| 3 | DeepSeek V4 Pro | 69.1% ±2.24 SE | $26.34 | 48.4 min | #1 same-model | Ante0.preview.54 | Harbor link | Jul 7, 2026 | |
| 4 | DeepSeek V4 Flash | 66.4% ±2.27 SE | $49.98 | 41.4 min | #1 same-model | Ante0.preview.53 | Harbor link | Jul 5, 2026 | |
| 5 | MiMo V2.5 | 65.8% ±2.30 SE | $73.76 | 12.5 min | #1 same-model | Ante20260625-0824-e383a92 | Harbor link | Jun 25, 2026 | |
| 6 | MiniMax M3 | 62.1% ±2.33 SE | $121.00 | 13.9 min | #1 same-model | Ante20260623-0825-ff174ee | Harbor link | Jun 23, 2026 | |
| 7 | Qwen3.6 27B | 56.2% ±2.36 SE | Local | 61.6 min | No public rows | Ante20260701-0837-70d2aac | Harbor link | Jul 3, 2026 |
GLM 5.2
DeepSeek V4 Pro
MiMo V2.5
MiniMax M3
Qwen3.6 27BTerminal-Bench Reference
Verified Public Leaderboard
Same parameters: 89 tasks · 5 trials/task · Updated Jun 21, 2026 · Source: Terminal-Bench official verified rows
13 official rowsFor how different models perform on TB 2.1, see Vals AI's Terminal-Bench 2.1 benchmark.
| # | Agent | Model | Accuracy | Run date | |
|---|---|---|---|---|---|
| 1 | Codex CLIOpenAI | GPT-5.5 | 83.4% ±1.11 SE | May 1, 2026 | |
| 2 | Claude CodeAnthropic | Claude 5 Fable | 83.2% ±1.04 SE | Jun 17, 2026 | |
| 3 | Terminus 2Terminal-Bench | Claude 5 Fable | 80.5% ±1.16 SE | Jun 17, 2026 | |
| 4 | Claude CodeAnthropic | Claude Opus 4.8 | 78.9% ±1.26 SE | May 29, 2026 | |
| 5 | Terminus 2Terminal-Bench | GPT-5.5 | 78.2% ±1.20 SE | May 1, 2026 | |
| 6 | Terminus 2Terminal-Bench | Claude Opus 4.8 | 74.6% ±1.23 SE | May 29, 2026 | |
| 7 | Terminus 2Terminal-Bench | Gemini 3 Pro | 74.4% ±1.32 SE | May 1, 2026 | |
| 8 | Gemini CLIGoogle | Gemini 3.1 Pro | 70.7% ±1.48 SE | May 5, 2026 | |
| 9 | Terminus 2Terminal-Bench | Gemini 3.1 Pro | 70.3% ±1.48 SE | May 5, 2026 | |
| 10 | Claude CodeAnthropic | Claude Opus 4.7 | 69.7% ±1.39 SE | May 1, 2026 | |
| 11 | Gemini CLIGoogle | Gemini 3 Pro | 66.3% ±1.37 SE | May 2, 2026 | |
| 12 | Terminus 2Terminal-Bench | Claude Opus 4.7 | 66.1% ±1.37 SE | May 1, 2026 | |
| 13 | Claude CodeAnthropic | GLM 5.1 | 58.6% ±1.24 SE | May 2, 2026 |
GPT-5.5
Claude 5 Fable
Gemini 3 Pro