Mathematics · Benchmark profile
Harvard-MIT Mathematics Tournament February 2026 problems used as a contest math evaluation.
Data verified 15 July 2026 · Methodology 1.2.0
Benchmark score on HMMT Feb 2026
Qwen3.7-Max leads at 97.1%, followed by Qwen3.7-Plus (92.9%) and GLM-5.2 (92.5%).
Visual analysis
Three views of the same sourced leaderboard: model placement, the score curve, and descriptive provider averages. These visuals never alter ranking calculations.
10 ranked models · higher is better
| Model | Rank | Provider | Score |
|---|---|---|---|
| Qwen3.7-Max | #1 | Alibaba Cloud | 97.1% |
| Qwen3.7-Plus | #2 | Alibaba Cloud | 92.9% |
| GLM-5.2 | #3 | Z.AI | 92.5% |
| Qwen3.6 Plus | #4 | Alibaba Cloud | 87.8% |
| Kimi K2.5 | #5 | Moonshot AI | 87.1% |
| GLM-5 | #6 | Z.AI | 86.4% |
| Claude Opus 4.5 | #7 | Anthropic | 85.3% |
| GLM-5.1 | #8 | Z.AI | 82.6% |
| DeepSeek V4 Flash | #9 | DeepSeek | 40.8% |
| DeepSeek V4 Pro | #10 | DeepSeek | 31.7% |
Shows how quickly scores separate down the field
| Model | Rank | Provider | Score |
|---|---|---|---|
| Qwen3.7-Max | #1 | Alibaba Cloud | 97.1% |
| Qwen3.7-Plus | #2 | Alibaba Cloud | 92.9% |
| GLM-5.2 | #3 | Z.AI | 92.5% |
| Qwen3.6 Plus | #4 | Alibaba Cloud | 87.8% |
| Kimi K2.5 | #5 | Moonshot AI | 87.1% |
| GLM-5 | #6 | Z.AI | 86.4% |
| Claude Opus 4.5 | #7 | Anthropic | 85.3% |
| GLM-5.1 | #8 | Z.AI | 82.6% |
| DeepSeek V4 Flash | #9 | DeepSeek | 40.8% |
| DeepSeek V4 Pro | #10 | DeepSeek | 31.7% |
Average within the visible cohort · descriptive only
| Provider | Models | Average score |
|---|---|---|
| Alibaba Cloud | 3 | 92.6% |
| Z.AI | 3 | 87.2% |
| Moonshot AI | 1 | 87.1% |
| Anthropic | 1 | 85.3% |
| DeepSeek | 2 | 36.3% |
One best score per model · higher is better
| Rank | Model | Provider | License | Score |
|---|---|---|---|---|
| #1 | Qwen3.7-Max qwen3.7-max | Alibaba Cloud | — | 97.1% |
| #2 | Qwen3.7-Plus qwen3.7-plus | Alibaba Cloud | — | 92.9% |
| #3 | GLM-5.2 glm-5-2 | Z.AI | open | 92.5% |
| #4 | Qwen3.6 Plus qwen3-6-plus | Alibaba Cloud | closed | 87.8% |
| #5 | Kimi K2.5 kimi-k2.5 | Moonshot AI | — | 87.1% |
| #6 | GLM-5 glm-5 | Z.AI | open | 86.4% |
| #7 | Claude Opus 4.5 claude-opus-4-5 | Anthropic | closed | 85.3% |
| #8 | GLM-5.1 glm-5-1 | Z.AI | open | 82.6% |
| #9 | DeepSeek V4 Flash deepseek-v4-flash | DeepSeek | open | 40.8% |
| #10 | DeepSeek V4 Pro deepseek-v4-pro | DeepSeek | open | 31.7% |
The top of this snapshot is led by Qwen3.7-Max at 97.1%; third place is 4.6 points behind. The top-10 spread is 65.4 points.
About HMMT Feb 2026
Harvard-MIT Mathematics Tournament February 2026 problems used as a contest math evaluation. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.
Open benchmark source ↗FAQ
Harvard-MIT Mathematics Tournament February 2026 problems used as a contest math evaluation.
Qwen3.7-Max by Alibaba Cloud currently leads with 97.1%.
10 models in the LuminaBench cohort have a qualifying score on this benchmark.
Yes. This family is ranking-weighted at 2.3% of its capability category. Qwen3.7-Max is overall #13.
Related
