Revelo Code Index - Terminal Bench 2.1
Code · 2026-07-29
The Revelo Code Index's Terminal Bench 2.1 task set: 20 private multi-step command-line tasks (building, debugging, configuration, data processing and system repair) written by hand into isolated containers in the Terminal-Bench 2.1 format with Terminal-Bench 2.0's domain proportions, though none are Terminal-Bench's own tasks. Verifiers check the final state of the container, so any route to it passes, and the score is the mean over tasks of three trials.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 68.3 |
| DeepSeek V4.1 Flash | 65.0 |
| GPT-6 Astra | 65.0 |
| GPT-5.6 Sol | 58.3 |
| Claude Fable 5 | 51.7 |
| Gemini 3.8 Flash | 50.0 |
| Kimi K3 | 50.0 |
| Opus 4.8 | 48.3 |
| GPT-5.6 Terra | 41.7 |
| Qwen3.8-Max | 41.7 |
| GPT-5.6 Luna | 40.0 |
| DeepSeek-V4-Flash-0731 | 38.3 |
| GLM-5.2 | 38.3 |
| Grok 4.5 | 35.0 |
| Gemini 3.6 Flash | 23.3 |
| DeepSeek-V4-Pro | 21.7 |
| Hy3 | 21.7 |
| MiniMax M3 | 20.0 |
| MiMo-V2.5-Pro | 8.3 |