# ARC-AGI-2 → HLE w/o tools

Calibration · https://benchgap.net/calibration/arc-agi-2/current/to/hlenotools/current · data retrieved 2026-11-21

HLE w/o tools is estimated from ARC-AGI-2 with a linear curve fitted on 18 models measured on both: y = 0.3391·x + 0.2112, R² = 0.66, cross-validated error 6.6 pp. It is used for 0 estimates.

## Fitted on

| Model | ARC-AGI-2 | HLE w/o tools |
|---|---|---|
| [Claude Fable 5.1](https://benchgap.net/model/claude-fable-5-1.md) | 90.0% | 60.9% |
| [Claude Haiku 5.5](https://benchgap.net/model/claude-haiku-5-5.md) | 61.3% | 45.9% |
| [Claude Opus 4.7](https://benchgap.net/model/claude-opus-4-7.md) | 75.8% | 46.9% |
| [Claude Opus 4.8](https://benchgap.net/model/claude-opus-4-8.md) | 72.1% | 49.8% |
| [Claude Opus 5](https://benchgap.net/model/claude-opus-5.md) | 90.4% | 56.3% |
| [Claude Opus 5.5](https://benchgap.net/model/claude-opus-5-5.md) | 91.7% | 64.4% |
| [Gemini 3.1 Pro](https://benchgap.net/model/gemini-3-1-pro.md) | 77.1% | 45.4% |
| [GPT-5.4](https://benchgap.net/model/gpt-5-4.md) | 74.0% | 39.8% |
| [GPT-5.4 mini](https://benchgap.net/model/gpt-5-4-mini.md) | 18.9% | 28.2% |
| [GPT-5.4 nano](https://benchgap.net/model/gpt-5-4-nano.md) | 5.7% | 24.3% |
| [GPT-5.4 Pro](https://benchgap.net/model/gpt-5-4-pro.md) | 83.3% | 42.7% |
| [GPT-5.5](https://benchgap.net/model/gpt-5-5.md) | 85.0% | 41.4% |
| [GPT-5.5 Pro](https://benchgap.net/model/gpt-5-5-pro.md) | 84.2% | 43.1% |
| [Grok 4.20](https://benchgap.net/model/grok-4-20-beta.md) | 53.3% | 31.6% |
| [Inkling 2 Small](https://benchgap.net/model/inkling-2-small.md) | 81.7% | 44.1% |
| [Inkling-Small](https://benchgap.net/model/inkling-small.md) | 40.1% | 31.6% |
| [Kimi K3](https://benchgap.net/model/kimi-k3.md) | 60.4% | 43.5% |
| [Muse Spark](https://benchgap.net/model/muse-spark.md) | 42.5% | 42.8% |
