Terminal-Bench 2.1 - Medium
Agents · 2026-05-06
The Medium split of Terminal-Bench 2.1 reports the percentage of medium-difficulty terminal-use tasks completed successfully; higher scores are better.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol | 86.1 |
| GPT-6 Astra | 86.1 |
| Gemini 3.8 Flash | 84.8 |
| GPT-6 Sol | 84.2 |
| Gemini 3.7 Flash | 83.6 |
| GPT-5.6 Luna | 83.6 |
| Grok 4.6 | 83.6 |
| Kimi K3 | 83.6 |
| DeepSeek V4.1 Flash | 83.0 |
| Opus 4.7 | 82.4 |
| Sonnet 5 | 81.8 |
| GPT-5.5 | 81.2 |
| Muse Spark 1.3 | 81.2 |
| Gemini 3.5 Flash | 80.0 |
| GLM-5.3 | 80.0 |
| DeepSeek-V4-Flash-0731 | 79.4 |
| GPT-5.6 Terra | 79.4 |
| Muse Spark 1.1 | 78.2 |
| Muse Spark 1.2 | 78.2 |
| Gemini 3.6 Flash | 77.0 |
| Kimi K2.7 Code | 77.0 |
| Opus 4.8 | 76.4 |
| GPT-6 Luna | 75.8 |
| Gemini 3.1 Pro Preview | 75.2 |
| Grok 4.7 | 75.2 |
| MiMo-V2.6-Flash | 75.2 |
| GLM-5.2 | 74.5 |
| Qwen3.8-Max | 74.5 |
| GLM-5.3 Flash | 73.9 |
| Grok 4.5 | 72.1 |
| MiMo-V2.5 | 70.9 |
| MiMo-V2.6-Pro | 69.7 |
| Qwen3.7-Max (source unspecified) | 68.5 |
| DeepSeek V4 Pro 0813 | 67.3 |
| Qwen3.7-Plus (source unspecified) | 67.3 |
| Qwen3.8 27B | 67.3 |
| Sonnet 4.6 | 66.7 |
| Inkling-Small | 66.7 |
| Kimi K2.6 | 66.7 |
| Composer 2.5 | 65.5 |