Which AI model leads HMMT February 2026?

GPT-5.5 leads HMMT February 2026 with 98.5% as of 25 Oct 2026, ahead of GPT-5.4 at 97.7%.

How many models have HMMT February 2026 results?

31 model configurations have a HMMT February 2026 result on BenchLeader, all taken from MathArena.

Who runs HMMT February 2026 and how often is it updated?

HMMT February 2026 is published by MathArena. BenchLeader re-reads the published results once a day and records the date each result was published.

Does HMMT February 2026 count toward the BenchLeader Index?

Yes. HMMT February 2026 contributes to the maths category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.

HMMT February 2026

Harvard-MIT Mathematics Tournament, February 2026. MathArena.

As of 25 Oct 2026, GPT-5.5 leads HMMT February 2026 on BenchLeader with 98.5%, ahead of GPT-5.4 at 97.7%, across 31 model configurations with a published result.

MathArena marks 21 of these 31 results as possibly contaminated: the model was evaluated after the questions were public, so they may already be in its training data. Those scores are in the table below, and count in the index at the same weight as the rest.

Published by: MathArena
Category: Maths
Index weight: 1.0
Models: 31
Data as of: 25 Oct 2026

MIT — eth-sri/matharena.

What the test looks like

The February 2026 Harvard-MIT Mathematics Tournament, a harder competition than AIME, evaluated on release by MathArena.

How it is scored

Percent of problems answered correctly, averaged over runs.

What to keep in mind

Small problem count and high difficulty make the score volatile; compare with confidence intervals on MathArena.

Top 15

31 of 31

#
1	GPT-5.5 xhighOpenAI	98.5%	66.1	possible
2	GPT-5.4 xhighOpenAI	97.7%	64.7	possible
3	GPT-5.2 highOpenAI	97.0%	56.0	–
4	Kimi K3 thinkingMoonshot AIopen ↗	97.0%	–	possible
5	Claude Opus 4.6 highAnthropic	96.2%	62.7	–
6	Claude Opus 4.8 maxAnthropic	95.5%	63.3	possible
7	Gemini 3.5 FlashGoogle	95.5%	48.8	possible
8	Gemini 3.1 ProGoogle	94.7%	60.5	possible
9	Kimi K2.6 thinkingMoonshot AIopen ↗	94.7%	59.4	possible
10	DeepSeek V4 Pro maxDeepSeekopen ↗	93.9%	60.8	possible
11	DeepSeek V4 Flash maxDeepSeekopen ↗	93.9%	59.4	possible
12	Claude Opus 4.7 xhighAnthropic	93.9%	57.4	possible
13	GLM 5.2Zhipu AIopen ↗	92.4%	51.8	possible
14	Gemini 3 FlashGoogle	89.4%	57.0	–
15	GLM 5.1Zhipu AIopen ↗	89.4%	55.9	possible
16	Gemini 3.6 FlashGoogle	89.4%	49.5	possible
17	Qwen3.5 397B A17BAlibabaopen ↗	87.9%	51.5	possible
18	Step 3.7 FlashStepFunopen ↗	87.9%	51.5	possible
19	Kimi K2.5 thinkingMoonshot AIopen ↗	87.1%	57.6	–
20	Gemini 3 ProGoogle	86.4%	60.4	–
21	GLM-5Zhipu AIopen ↗	86.4%	54.9	–
22	Grok 4.1 Fast thinkingSpaceXAI	86.4%	54.7	–
23	Step 3.5 FlashStepFunopen ↗	86.4%	53.7	–
24	Nemotron 3 SuperNVIDIA	84.8%	44.3	possible
25	Qwen3.5-35B-A3BAlibabaopen ↗	81.8%	50.2	possible
26	Qwen3.5-27BAlibabaopen ↗	81.1%	52.3	possible
27	Qwen3 30B A3B 2507 thinkingAlibabaopen ↗	78.8%	43.2	–
28	Qwen3.5-4BAlibabaopen ↗	72.9%	–	possible
29	Qwen3.5-9BAlibabaopen ↗	71.2%	40.1	possible
30	Qed NanoLM-Provers	64.4%	–	possible
31	Qwen3-4B-Instruct-2507 thinkingAlibabaopen	53.0%	40.0	–

More maths benchmarks · All LLM benchmarks

Cite as: BenchLeader, “HMMT February 2026 leaderboard”, data as of 25 Oct 2026, https://www.benchleader.com/benchmarks/matharena_hmmt_feb_2026.

HMMT February 2026: questions

What does HMMT February 2026 measure?: The February 2026 Harvard-MIT Mathematics Tournament, a harder competition than AIME, evaluated on release by MathArena. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads HMMT February 2026?: GPT-5.5 leads HMMT February 2026 with 98.5% as of 25 Oct 2026, ahead of GPT-5.4 at 97.7%.
How many models have HMMT February 2026 results?: 31 model configurations have a HMMT February 2026 result on BenchLeader, all taken from MathArena.
Who runs HMMT February 2026 and how often is it updated?: HMMT February 2026 is published by MathArena. BenchLeader re-reads the published results once a day and records the date each result was published.
Does HMMT February 2026 count toward the BenchLeader Index?: Yes. HMMT February 2026 contributes to the maths category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.