LLM Stats (Terminal-Bench 2.1): leaderboard
LLM Stats mirror of Terminal-Bench 2.1, measuring agentic terminal use on software engineering and system tasks.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Estimated already saturated. 46 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4.1 Flash | 90.6 | #36 |
| 2 | MiMo-V2.6-Pro | 89.9 | #42 |
| 3 | Gemini 3.8 Flash | 89.4 | #21 |
| 4 | GPT-5.6 Sol | 88.8 | #19 |
| 5 | Muse Spark 1.3 | 88.8 | #24 |
| 6 | Kimi K3 | 88.3 | #30 |
| 7 | GLM-5.3 | 88.2 | #39 |
| 8 | DeepSeek V4 Pro (0813) | 87.9 | #53 |
| 9 | MiMo-V2.6-Flash | 87.6 | #70 |
| 10 | GPT-5.6 Terra | 87.4 | #41 |
| 11 | Qwen 3.8 Max | 86.6 | #35 |
| 12 | Gemini 3.7 Flash | 85.8 | #23 |
| 13 | Hy4 preview | 85.4 | #38 |
| 14 | GPT-5.6 Luna | 84.7 | #68 |
| 15 | GLM-5.3 Flash | 84.3 | #65 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-terminal-bench-2-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.