// BENCHMARK

HMMT February 2025 benchmark

AI model leaderboard for the HMMT February 2025 benchmark. Compare how large language models score on HMMT February 2025, see the full ranking, and understand what this AI benchmark measures. Gemini 3 Pro currently leads with 97.5%. Harvard-MIT Mathematics Tournament February 2025 — 30 algebra/number-theory/combinatorics/geometry questions.

Leaderboard

# Model Organization Score Variant Source
#1 Gemini 3 Pro Google DeepMind 97.5% cited-deepseek-v3-2 official ↗
#2 Kimi K2 Thinking Moonshot 97.5% thinking-heavy official ↗
#3 Kimi K2 Thinking Moonshot 95.1% thinking-w-python official ↗
#4 GPT-5 OpenAI 93.3% thinking-high-cited-ring-1t official ↗
#5 DeepSeek-V3.2 DeepSeek 92.5% thinking official ↗
#6 GPT-5 OpenAI 91.79% High official ↗
#7 Kimi K2 Thinking Moonshot 89.4% thinking-no-tools official ↗
#8 Kimi K2 Thinking Moonshot 89.4% thinking-cited-deepseek-v3-2 official ↗
#9 GPT-5 OpenAI 88.3% cited-deepseek-v3-2 official ↗
#10 Ring-1T Ant Group 86.72% thinking-avg@16 official ↗
#11 DeepSeek-V3.1-Terminus DeepSeek 86.1% thinking-cited-ring-1t official ↗
#12 Qwen3-235B-A22B-Thinking (Jul 2025) Qwen 83.9% thinking-cited-ring-1t official ↗
#13 DeepSeek-V3.2-Exp DeepSeek 83.6% from-hf-readme official ↗
#14 DeepSeek-V3.1 DeepSeek 83.54% official ↗
#15 gemini-2.5-pro Google DeepMind 82.5% cited-phi4r official ↗
#16 gemini-2.5-pro Google DeepMind 82.5% cited-ring-1t official ↗
#17 gpt-oss-120b OpenAI 81.88% official ↗
#18 o3 OpenAI 80.8% High official ↗
#19 DeepSeek-R1-0528 DeepSeek 79.4% reasoning official ↗
#20 Claude Sonnet 4.5 Anthropic 79.2% cited-deepseek-v3-2 official ↗
#21 gemini-2.5-pro Google DeepMind 74.18% official ↗
#22 K2 Think Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) 73.75% avg@16 official ↗
#23 EXAONE 4.0 (32B) LG AI Research 72.9% reasoning official ↗
#24 Kimi K2 Moonshot 70.4% k2-0905-w-python-cited-k2-thinking official ↗
#25 gpt-oss-20b OpenAI 69.38% official ↗
#26 o3-mini OpenAI 67.5% high-cited-phi4r official ↗
#27 OpenReasoning-Nemotron-32B NVIDIA 67.29% official ↗
#28 Qwen3-235B-A22B-Thinking (Jul 2025) Qwen 62.5% reasoning official ↗
#29 Qwen3-235B-A22B Qwen 61.88% official ↗
#30 gemini-2.5-pro Google DeepMind 60.73% lowthink-cited-ling-1t official ↗
#31 Phi-4-Reasoning-plus Microsoft 53.6% reasoning official ↗
#32 Phi-4-Reasoning-plus Microsoft 53.6% pass@1-avg@64 official ↗
#33 QwQ-32B Qwen 51.46% official ↗
#34 Qwen3-32B Qwen 50.4% reasoning official ↗
#35 o1 OpenAI 48.3% cited-phi4r official ↗
#36 QwQ-32B Qwen 47.5% cited-phi4r official ↗
#37 Ling-1T Ant Group 47.08% pass@1 official ↗
#38 DeepSeek-R1 DeepSeek 47.08% official ↗
#39 Phi-4-Reasoning Microsoft 43.8% pass@1-avg@64 official ↗
#40 DeepSeek-R1 DeepSeek 41.7% cited-phi4r official ↗
#41 DeepSeek-V3.1-Terminus DeepSeek 41.25% non-thinking-cited-ling-1t official ↗
#42 Kimi K2 Moonshot 38.8% instruct-0905-cited-ling-1t official ↗
#43 Kimi K2 Moonshot 38.8% k2-0905-no-tools-cited-k2-thinking official ↗
#44 o1-mini OpenAI 38% cited-phi4r official ↗
#45 GPT-5 OpenAI 36.98% main-cited-ling-1t official ↗
#46 EXAONE 4.0 (1.2B) LG AI Research 34% reasoning official ↗
#47 DeepSeek-R1-Distill-Llama-70B DeepSeek 33.3% cited-phi4r official ↗
#48 Claude 3.7 Sonnet Anthropic 31.7% thinking-cited-phi4r official ↗
#49 Qwen3-30B-A3B Qwen 23.54% official ↗
#50 Kimi Linear Moonshot 12.5% sft-avg@32 official ↗

Frequently asked questions about HMMT February 2025

What is the HMMT February 2025 benchmark?

Harvard-MIT Mathematics Tournament February 2025 — 30 algebra/number-theory/combinatorics/geometry questions.

How is the HMMT February 2025 benchmark scored?

HMMT February 2025 is scored using the accuracy (%) metric, where a higher score is better. The maximum achievable score is 100.000. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.

Which AI model scores highest on HMMT February 2025?

As of the latest reported scores on GenAIList, Gemini 3 Pro achieves the highest result on HMMT February 2025 with a score of 97.5%.

Is a higher HMMT February 2025 score better?

Yes. On HMMT February 2025 a higher score indicates better performance, so models near the top of the leaderboard are the strongest.