HMMT Feb 26
HMMT Feb 26 - Harvard-MIT Mathematics Tournament February 2026.
Category
STEM & Reasoning
Score type
percent
Scale
100%
higher is better
Models tested
31
What is HMMT Feb 26?
HMMT Feb 26 is the February round of the Harvard-MIT Mathematics Tournament 2026. HMMT problems are considerably harder than AIME and count among the most difficult high-school competition math in the USA. Models must solve the problems with exact answers — recent tournaments are almost immune to training-data leakage.
What does it test?
HMMT Feb 26 measures hard reasoning in science, technology, engineering and math. In detail, it checks:
- olympiade-level insight instead of memorising standard tasks
- correct exact answers within a fixed time window
- advanced algebra, combinatorics, number theory and geometry
- resistance to data contamination thanks to very recent problems
How to read the score
The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.
Frequently asked questions
What is the HMMT Feb 26 benchmark?
HMMT Feb 26 is the February round of the Harvard-MIT Mathematics Tournament 2026. HMMT problems are considerably harder than AIME and count among the most difficult high-school competition math in the USA. Models must solve the problems with exact answers — recent tournaments are almost immune to training-data leakage.
What does HMMT Feb 26 test?
HMMT Feb 26 mainly measures hard reasoning in science, technology, engineering and math. Concretely, it checks: olympiade-level insight instead of memorising standard tasks; correct exact answers within a fixed time window; advanced algebra, combinatorics, number theory and geometry.
How is the HMMT Feb 26 score calculated?
The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.
Where do I get HMMT Feb 26 data and scores?
The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/hmmt-feb-26/scores) or pull it directly from the Open LLM Radar API with any HTTP client.
Benchmark data & downloads
Where to get, inspect and reuse the data behind this benchmark.
Benchmark dataset
The question/task data itself — download and evaluate locally
Benchmark metadata (JSON)
This page's benchmark record from the Open LLM Radar API
All model scores (JSON download)
Full score table as JSON; paginate with limit/offset params
Open LLM Radar API docs
Swagger UI to explore and reuse all radar API endpoints
Model scores 31 models tested
| # | Model | Score | Date | Source | Verified |
|---|---|---|---|---|---|
| 1 |
Qwen3.8-27B
Qwen Team (Alibaba Cloud)
|
95.8%
|
21.10.2026 | Report | Unverified |
| 2 |
DeepSeek-V4-Pro
DeepSeek
|
95.2%
|
17.06.2026 | Report | Unverified |
| 3 |
DeepSeek-V4-Pro
DeepSeek
|
95.2%
|
26.06.2026 | Report | Verified |
| 4 |
DeepSeek-V4-Flash
DeepSeek
|
94.8%
|
26.06.2026 | Report | Verified |
| 5 |
Muse-Spark-Open-816B-A42B
Meta Superintelligence Lab
|
93.9%
|
— | Report | Unverified |
| 6 |
Kimi K2.6
Moonshot AI
|
92.7%
|
23.08.2026 | Report | Verified |
| 7 |
GLM-5.2
Zhipu AI
|
92.5%
|
17.06.2026 | Report | Verified |
| 8 |
Ripple-65B Preview
Ricursive Intelligence
|
90.5%
|
— | Report | Verified |
| 9 |
Kimi K2.5
Moonshot AI
|
87.1%
|
23.08.2026 | Report | Verified |
| 10 |
MiniMax-M3
MiniMaxAI
|
84.4%
|
17.06.2026 | Report | Unverified |
| 11 |
Qwen3.6-27B
Qwen Team (Alibaba Cloud)
|
84.3%
|
22.04.2026 | Qwen Team (official model card) | Verified |
| 12 |
Qwen3.5-27B
Qwen Team (Alibaba Cloud)
|
84.3%
|
16.04.2026 | Qwen Team (official model card - comparison) | Unverified |
| 13 |
Raon-1 mini
KRAFTON AI
|
84.1%
|
21.10.2026 | Report | Verified |
| 14 |
Qwen3.6-35B-A3B
Qwen Team (Alibaba Cloud)
|
83.6%
|
16.04.2026 | Qwen Team (official model card); KRAFTON re-eval (Raon-1-mini-preview card, gpt-5.6-luna harness): 84.5 | Verified |
| 15 |
Spark-X2.5-4B
SparkLLM (XHToken)
|
81.2%
|
— | Report | Verified |
| 16 |
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16
NVIDIA
|
80.3%
|
— | Report | Unverified |
| 17 |
Gemma4-26B-A4B
Google
|
79%
|
16.04.2026 | Qwen Team (official model card - comparison); KRAFTON re-eval (Raon-1-mini-preview card, gpt-5.6-luna harness): 79.9 | Unverified |
| 18 |
Qwen3.5-35B-A3B
Qwen Team (Alibaba Cloud)
|
78.7%
|
16.04.2026 | Qwen Team (official model card - comparison) | Unverified |
| 19 |
Gemma4-31B-it
Google
|
77.2%
|
16.04.2026 | Qwen Team (official model card - comparison) | Unverified |
| 20 |
Muse-Glimmer-30B
Meta Superintelligence Lab
|
74.6%
|
21.10.2026 | Report | Unverified |
| 21 |
Qwen3.5-4B
Qwen
|
69.7%
|
— | Report | Verified |
| 22 |
Gemma4-12B
Google
|
65.6%
|
— | Report | Verified |
| 23 |
MiniCPM5-2B
OpenBMB
|
63.8%
|
— | Report | Verified |
| 24 |
granite-4.2-3B
IBM
|
60.8%
|
— | Report | Verified |
| 25 |
Nemotron-3-Nano-4B
NVIDIA
|
51.3%
|
— | Report | Verified |