HMMT 2025 — Python-tool-enabled configuration, distinct scores from no-tools and heavy tiers.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | GPT-5 | OpenAI | 96.7 | 1 | 2026-05-15 |
| 2 | Kimi K2 (Reasoning) | Moonshot | 95.1 | 2 | 2026-05-15 |
| 3 | Grok 4 | SpaceXAI | 93.9 | 1 | 2026-05-15 |
| 4 | Claude Sonnet 4.5 | Anthropic | 88.8 | 2 | 2026-06-12 |
| 5 | DeepSeek-V3.2 | DeepSeek | 49.5 | 1 | 2026-05-15 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.