# Terminal-Bench 2.0 → Terminal-Bench Hard

Calibration · https://benchgap.net/calibration/1498 · data retrieved 2026-10-29

Terminal-Bench Hard is estimated from Terminal-Bench 2.0 with a Hill curve fitted on 12 models measured on both: y = 0.4107 + (1.2000 − 0.4107)·x^6.00 / (0.99545^6.00 + x^6.00), R² = 0.62, cross-validated error 4.0 pp. It is used for 18 estimates.

| Estimated model | Terminal-Bench 2.0 | Terminal-Bench Hard | Source |
|---|---|---|---|
| [GPT-5.3 Codex](https://benchgap.net/model/gpt-5-3-codex.md) | 77.3% | 55.3% | estimated ± 4.0 pp, high confidence |
| [Claude Opus 4.7 (Adaptive)](https://benchgap.net/model/claude-opus-4-7-adaptive.md) | 69.4% | 49.2% | estimated ± 4.0 pp, high confidence |
| [Composer 2](https://benchgap.net/model/composer-2.md) | 61.7% | 45.3% | estimated ± 4.0 pp, high confidence |
| [Claude Opus 4.5](https://benchgap.net/model/claude-opus-4-5.md) | 59.3% | 44.4% | estimated ± 4.0 pp, high confidence |
| [Qwen3.6-27B](https://benchgap.net/model/qwen3-6-27b.md) | 59.3% | 44.4% | estimated ± 4.0 pp, high confidence |
| [Hy3 Preview](https://benchgap.net/model/hy3-preview.md) | 54.4% | 43.1% | estimated ± 4.0 pp, medium confidence |
| [Qwen3.5 397B](https://benchgap.net/model/qwen3-5-397b.md) | 52.5% | 42.7% | estimated ± 4.0 pp, medium confidence |
| [Qwen3.6-35B-A3B](https://benchgap.net/model/qwen3-6-35b-a3b.md) | 51.5% | 42.6% | estimated ± 4.0 pp, medium confidence |
| [Kimi K2.5](https://benchgap.net/model/kimi-k2-5.md) | 50.8% | 42.4% | estimated ± 4.0 pp, medium confidence |
| [Kimi K2.5 (Reasoning)](https://benchgap.net/model/kimi-k2-5-reasoning.md) | 50.8% | 42.4% | estimated ± 4.0 pp, medium confidence |
| [Claude Sonnet 4.5](https://benchgap.net/model/claude-sonnet-4-5.md) | 50.0% | 42.3% | estimated ± 4.0 pp, medium confidence |
| [Qwen3.5-122B-A10B](https://benchgap.net/model/qwen3-5-122b-a10b.md) | 49.4% | 42.2% | estimated ± 4.0 pp, medium confidence |
| [MAI-Thinking-1](https://benchgap.net/model/mai-thinking-1.md) | 46.0% | 41.8% | estimated ± 4.0 pp, medium confidence |
| [Qwen3.5-27B](https://benchgap.net/model/qwen3-5-27b.md) | 41.6% | 41.5% | estimated ± 4.0 pp, medium confidence |
| [GLM-4.7](https://benchgap.net/model/glm-4-7.md) | 41.0% | 41.5% | estimated ± 4.0 pp, medium confidence |
| [Qwen3.5-35B-A3B](https://benchgap.net/model/qwen3-5-35b-a3b.md) | 40.5% | 41.4% | estimated ± 4.0 pp, medium confidence |
| [Laguna XS 2.1](https://benchgap.net/model/laguna-xs-2-1.md) | 37.5% | 41.3% | estimated ± 4.0 pp, medium confidence |
| [LongCat-Flash-Lite-Sparse](https://benchgap.net/model/longcat-flash-lite-sparse.md) | 33.7% | 41.2% | estimated ± 4.0 pp, medium confidence |

## Fitted on

| Model | Terminal-Bench 2.0 | Terminal-Bench Hard |
|---|---|---|
| [Claude Opus 4.6](https://benchgap.net/model/claude-opus-4-6.md) | 65.4% | 48.5% |
| [GLM-5](https://benchgap.net/model/glm-5.md) | 56.2% | 43.2% |
| [GPT-5.4](https://benchgap.net/model/gpt-5-4.md) | 75.1% | 57.6% |
| [GPT-5.4 mini](https://benchgap.net/model/gpt-5-4-mini.md) | 60.0% | 52.3% |
| [GPT-5.5](https://benchgap.net/model/gpt-5-5.md) | 82.0% | 60.6% |
| [Kimi K2.6](https://benchgap.net/model/kimi-2-6.md) | 66.7% | 43.9% |
| [MiMo-V2.5-Pro](https://benchgap.net/model/mimo-v2-5-pro.md) | 68.4% | 43.2% |
| [Muse Spark](https://benchgap.net/model/muse-spark.md) | 59.0% | 45.5% |
| [Qwen 3.6 Max (preview)](https://benchgap.net/model/qwen3-6-max-preview.md) | 65.4% | 43.9% |
| [Qwen3.6 Plus](https://benchgap.net/model/qwen3-6-plus.md) | 61.6% | 43.9% |
| [Qwen3.7 Max](https://benchgap.net/model/qwen3-7-max.md) | 69.7% | 50.8% |
| [Qwen3.7 Plus](https://benchgap.net/model/qwen3-7-plus.md) | 70.3% | 47.0% |