Terminal-Bench 4.0 — Security
Agents · 2026-09-22
Resolution rate on the five security tasks in Terminal-Bench 4.0.
Top models (higher is better)
| Model | Score |
|---|---|
| GLM-5.3 Flash | 93.3 |
| GPT-6.1 Sol | 93.3 |
| Qwen3.8 Max (0902) | 93.3 |
| Muse Wonder Full | 86.7 |
| GPT-6.1 Terra | 80.0 |
| Gemini 4 Argon | 73.3 |
| Gemini 4 Carbon | 73.3 |
| GPT-6 Astra | 73.3 |
| Kimi K3.1 Pro | 73.3 |
| Step 5 Preview | 73.3 |
| Muse Wonder | 66.7 |
| GLM-5.3 | 60.0 |
| GPT-6 Sol | 60.0 |
| GPT-5.6 Sol | 53.3 |
| Grok 4.7 | 53.3 |
| GPT-5.6 Terra | 46.7 |
| Mistral Large 4 Preview | 46.7 |
| Muse Spark 1.3 | 46.7 |
| Claude Haiku 5.5 | 40.0 |
| Ember-1 | 40.0 |
| Opus 4.8 | 33.3 |
| Gemini 3.7 Flash | 33.3 |
| Gemini 3.8 Flash | 33.3 |
| MiMo-V2.6-Flash | 33.3 |
| MiniMax M3.1 Flash | 33.3 |
| DeepSeek V4.1 Flash 1017 | 26.7 |
| GPT-6 Luna | 26.7 |
| Grok 4.6 | 26.7 |
| Kimi K3 | 26.7 |
| MiMo-V2.6-Pro | 26.7 |
| Grok 4.5 | 20.0 |
| Hy4 Preview | 20.0 |
| Qwen3.8 27B | 13.3 |
| DeepSeek V4.1 Flash | 6.7 |
| DeepSeek-V4-Flash-0731 | 6.7 |
| DeepSeek V4 Pro 0813 | 6.7 |
| GPT-5.6 Luna | 6.7 |
| Mercury 3 | 6.7 |
| Muse Spark Open | 6.7 |
| Gemini 3.6 Flash | 6.7 |