# τ²-bench results → HLE w/ tools

Calibration · https://benchgap.net/calibration/1493 · data retrieved 2026-10-29

HLE w/ tools is estimated from τ²-bench results with a Hill curve fitted on 8 models measured on both: y = 0.0000 + (0.7350 − 0.0000)·x^6.00 / (0.83565^6.00 + x^6.00), R² = 0.67, cross-validated error 5.5 pp. It is used for 30 estimates.

| Estimated model | τ²-bench results | HLE w/ tools | Source |
|---|---|---|---|
| [GLM-5-Turbo](https://benchgap.net/model/glm-5-turbo.md) | 98.5% | 53.5% | estimated ± 5.5 pp, medium confidence |
| [GLM-5V-Turbo](https://benchgap.net/model/glm-5v-turbo.md) | 98.5% | 53.5% | estimated ± 5.5 pp, medium confidence |
| [Qwen 3.6 Max (preview)](https://benchgap.net/model/qwen3-6-max-preview.md) | 95.9% | 51.1% | estimated ± 5.5 pp, medium confidence |
| [MiMo-V2-Pro](https://benchgap.net/model/mimo-v2-pro.md) | 95.0% | 50.2% | estimated ± 5.5 pp, medium confidence |
| [Qwen3.5-27B](https://benchgap.net/model/qwen3-5-27b.md) | 93.9% | 49.1% | estimated ± 5.5 pp, medium confidence |
| [Grok 4.1 Fast (Reasoning)](https://benchgap.net/model/grok-4-1-fast-reasoning.md) | 93.3% | 48.5% | estimated ± 5.5 pp, medium confidence |
| [GPT-5.2-Codex](https://benchgap.net/model/gpt-5-2-codex.md) | 92.1% | 47.2% | estimated ± 5.5 pp, medium confidence |
| [Qwen3.5-4B](https://benchgap.net/model/qwen3-5-4b.md) | 92.1% | 47.2% | estimated ± 5.5 pp, medium confidence |
| [MiMo-V2-Omni](https://benchgap.net/model/mimo-v2-omni.md) | 91.2% | 46.2% | estimated ± 5.5 pp, medium confidence |
| [Claude Opus 4.5 Thinking](https://benchgap.net/model/claude-opus-4-5-thinking.md) | 89.5% | 44.2% | estimated ± 5.5 pp, medium confidence |
| [Qwen3.5-35B-A3B](https://benchgap.net/model/qwen3-5-35b-a3b.md) | 89.2% | 43.9% | estimated ± 5.5 pp, medium confidence |
| [LFM2.5-8B-A1B](https://benchgap.net/model/lfm2-5-8b-a1b.md) | 88.1% | 42.5% | estimated ± 5.5 pp, medium confidence |
| [Gemini 3 Pro](https://benchgap.net/model/gemini-3-pro.md) | 87.1% | 41.3% | estimated ± 5.5 pp, medium confidence |
| [GPT-5 (medium)](https://benchgap.net/model/gpt-5-medium.md) | 86.5% | 40.5% | estimated ± 5.5 pp, medium confidence |
| [GPT-5.3 Codex](https://benchgap.net/model/gpt-5-3-codex.md) | 86.0% | 39.9% | estimated ± 5.5 pp, medium confidence |
| [GPT-5.2](https://benchgap.net/model/gpt-5-2.md) | 84.8% | 38.4% | estimated ± 5.5 pp, medium confidence |
| [GPT-5.1-Codex](https://benchgap.net/model/gpt-5-1-codex.md) | 83.0% | 36.0% | estimated ± 5.5 pp, low confidence |
| [GPT-5.1-Codex-Max](https://benchgap.net/model/gpt-5-1-codex-max.md) | 83.0% | 36.0% | estimated ± 5.5 pp, low confidence |
| [o3](https://benchgap.net/model/o3.md) | 80.7% | 32.9% | estimated ± 5.5 pp, low confidence |
| [LLaDA2.2-flash](https://benchgap.net/model/llada2-2-flash.md) | 80.3% | 32.4% | estimated ± 5.5 pp, low confidence |
| [Ternary Bonsai 2 27B](https://benchgap.net/model/ternary-bonsai-2-27b.md) | 80.2% | 32.3% | estimated ± 5.5 pp, low confidence |
| [DeepSeek V3.2](https://benchgap.net/model/deepseek-v3-2.md) | 78.9% | 30.5% | estimated ± 5.5 pp, low confidence |
| [Agents-A1-4B](https://benchgap.net/model/agents-a1-4b.md) | 78.2% | 29.5% | estimated ± 5.5 pp, low confidence |
| [GLM-4.6](https://benchgap.net/model/glm-4-6.md) | 76.9% | 27.8% | estimated ± 5.5 pp, low confidence |
| [Grok Code Fast 1](https://benchgap.net/model/grok-code-fast-1.md) | 75.7% | 26.2% | estimated ± 5.5 pp, low confidence |
| [Grok 4](https://benchgap.net/model/grok-4.md) | 74.9% | 25.1% | estimated ± 5.5 pp, low confidence |
| [Qwen3 Max](https://benchgap.net/model/qwen3-max.md) | 74.3% | 24.3% | estimated ± 5.5 pp, low confidence |
| [Claude Opus 4.7](https://benchgap.net/model/claude-opus-4-7.md) | 74.0% | 23.9% | estimated ± 5.5 pp, low confidence |
| [Claude 4 Opus Thinking](https://benchgap.net/model/claude-4-opus-thinking.md) | 73.4% | 23.1% | estimated ± 5.5 pp, low confidence |
| [Claude 4.1 Opus Thinking](https://benchgap.net/model/claude-4-1-opus-thinking.md) | 71.4% | 20.6% | estimated ± 5.5 pp, low confidence |

## Fitted on

| Model | τ²-bench results | HLE w/ tools |
|---|---|---|
| [DeepSeek V4 Pro 0813](https://benchgap.net/model/deepseek-v4-pro-0813.md) | 96.2% | 60.0% |
| [GLM-5](https://benchgap.net/model/glm-5.md) | 98.2% | 50.4% |
| [GLM-5.1](https://benchgap.net/model/glm-5-1.md) | 97.7% | 52.3% |
| [GLM-5.2](https://benchgap.net/model/glm-5-2.md) | 99.1% | 54.7% |
| [Nemotron 3 Ultra](https://benchgap.net/model/nemotron-3-ultra.md) | 83.3% | 37.4% |
| [Qwen3.7 Max](https://benchgap.net/model/qwen3-7-max.md) | 94.7% | 53.5% |
| [Spark-X2.6-4B](https://benchgap.net/model/spark-x2-6-4b.md) | 86.9% | 36.9% |
| [Step 3.7 Flash](https://benchgap.net/model/step-3-7-flash.md) | 98.5% | 47.2% |