Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Mathematics benchmark
Harvard-MIT Mathematics Tournament February 2025. Every model the catalog carries a published HMMT Feb 2025 value for, ranked by that value.
A February 2025 HMMT slice used in exact-value provider tables for advanced contest-math reasoning. 16 models in the ModelChoice catalog have a published value.
Each point is a model with a published HMMT Feb 2025 value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
16 models with a published HMMT Feb 2025 value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Contest mathematics |
|---|---|---|---|
| 1 | Z.AI | 97.5 | |
| 2 | Z.AI | 97.1 | |
| 3 | Alibaba | 96.7 | |
What the benchmark measures and how its value is published, from the source’s own record.
HMMT Feb 2025 (Harvard-MIT Mathematics Tournament February 2025) is a mathematics benchmark published by Qwen3.6 launch benchmarks. A February 2025 HMMT slice used in exact-value provider tables for advanced contest-math reasoning. Its evaluation set is described by the source as competition math problems. The source rates its difficulty as olympiad-style mathematics.
Each model's result is published by Qwen3.6 launch benchmarks as contest mathematics. ModelChoice prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 16 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
| 4 |
| Moonshot AI |
| 95.4 |
| 4 | Moonshot AI | 95.4 |
| 6 | Alibaba | 94.8 |
| 7 | Alibaba | 93.8 |
| 8 | NVIDIA | 93.7 |
| 9 | Anthropic | 92.9 |
| 10 | Alibaba | 90.7 |
| 11 | IBM | 89.2 |
| 12 | SASarvam 105B | Sarvam | 85.8 |
| 13 | IBM | 78.3 |
| 14 | SASarvam 30B | Sarvam | 73.3 |
| 15 | IBM | 66.7 |
| 16 | DeepSeek | 61.5 |
Evidence key: ObservedLast good
Rows are ordered by the value Qwen3.6 launch benchmarks published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.