ToolMATH (Gold-present): leaderboard
Metric: Accuracy (%) when the model is given the gold tools plus five same-category distractor tools (similarity level 3); ToolMATH main split (about 8,000 MATH problems whose step-wise solutions are turned into hidden-code Python tools with descriptions and typed schemas); Plan+ReAct protocol, the model may also answer without calling a tool; exact-match answer accuracy after normalization; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 99.2 | #91 |
| 2 | Gemini 3.1 Pro (Preview) | 99 | #54 |
| 3 | GPT-4o Mini | 55.3 | #588 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=toolmath-gold-present · How It Works · Data refreshed daily, snapshot 2026-10-11.