Skip to main content
BenchLM
Data

AIME 2026 (AIME26)

Data verified

A 2026 American Invitational Mathematics Examination snapshot used in frontier-model comparison tables for mathematical reasoning.

Top models on AIME26 — November 21, 2026

As of November 21, 2026, GLM-5.2 leads the AIME26 leaderboard with 99.2% , followed by Beam 1.0 501B-A23B FP8 (98.0%) and Beam 1.0 501B-A23B (97.7%).

34 modelsMathematics6% of Math reference weightCurrentUpdated November 21, 2026

Leaderboard (34 models)

None of the 34 results on this page has a documented run by the benchmark or an independent evaluator: 31 reported by the labs that built the models and 3 reported in another publisher’s document.

AIME26 leaderboard
RankModel / configurationScoreParameters (B)Open / closed
1GLM-5.2Z.AIReported by Z.AI
99.2%
Not reportedOpen
2Beam 1.0 501B-A23B FP8Reflection AIReported by Reflection AI
98.0%
Not reportedOpen
3Beam 1.0 501B-A23BReflection AIReported by Reflection AI
97.7%
Not reportedOpen
4Beam 1.0 501B-A23B NVFP4Reflection AIReported by Reflection AI
97.5%
Not reportedOpen
5InklingThinking Machines LabReported by Thinking Machines Lab
97.1%
Not reportedOpen
6A.X K2SK TelecomReported by SK Telecom
97.1%
Not reportedOpen
7Kimi K2.6Moonshot AIReported by Moonshot AI
96.4%
Not reportedOpen
8Ternary Bonsai 2 27BPrism MLReported by Prism ML
95.8%
Not reportedOpen
9GLM-5Z.AIAnother lab’s comparison table
95.8%
Not reportedOpen
10Kimi K2.5Moonshot AIReported by Moonshot AI
95.8%
Not reportedOpen
11Solar Open 2UpstageReported by Upstage
95.7%
Not reportedOpen
12Inkling-SmallThinking Machines LabReported by Thinking Machines Lab
95.5%
Not reportedOpen
13Qwen3.6 PlusAlibabaReported by Alibaba
95.3%
Not reportedClosed
14GLM-5.1Z.AIReported by Z.AI
95.3%
Not reportedOpen
15Solar Pro 4UpstageReported by Upstage
95.3%
Not reportedClosed
16Claude Opus 4.5AnthropicAnother lab’s comparison table
95.1%
Not reportedClosed
17Ripple-65B PreviewRicursive IntelligenceReported by Ricursive Intelligence
95.0%
Not reportedOpen
18Muse Glimmer 30BMetaReported by Meta
94.7%
Not reportedOpen
19MAI-Thinking-1MicrosoftReported by Microsoft
94.5%
Not reportedClosed
20Qwen3.6-27BAlibabaReported by Alibaba
94.1%
Not reportedOpen
21Qwen3.5 397BAlibabaReported by Alibaba
93.3%
Not reportedOpen
22Ling 3.0 FlashInclusionAIReported by InclusionAI
93.2%
Not reportedOpen
23Qwen3.6-35B-A3BAlibabaReported by Alibaba
92.7%
Not reportedOpen
24K-EXAONE 2.0LG AI ResearchReported by LG AI Research
92.3%
Not reportedOpen
25Spark-X2.6-4BSparkLLMReported by SparkLLM
91.2%
Not reportedOpen
26MiniMax M2.7MiniMaxAnother lab’s comparison table
89.8%
Not reportedOpen
27ZAYA1-8BZyphraReported by Zyphra
89.1%
Not reportedOpen
28MiniCPM5-2BOpenBMBReported by OpenBMB
86.5%
Not reportedOpen
29Gemma 4 12BGoogleReported by Google
77.5%
Not reportedOpen
30ZAYA1-74B-PreviewZyphraReported by Zyphra
76.4%
Not reportedOpen
31LongCat-Flash-Lite-SparseMeituanReported by Meituan
65.7%
Not reportedOpen
32LFM2.5-8B-A1BLiquidAIReported by LiquidAI
50.0%
Not reportedOpen
33MiniCPM5-1BOpenBMBReported by OpenBMB
40.4%
Not reportedOpen
34LLaDA2.2-miniInclusionAIReported by InclusionAI
35.0%
Not reportedOpen
  • Lab-reported
  • Benchmark-run
  • Independent
What the labels mean

According to BenchLM.ai, GLM-5.2 leads the AIME26 benchmark with a score of 99.2%, followed by Beam 1.0 501B-A23B FP8 (98.0%) and Beam 1.0 501B-A23B (97.7%). The top models are clustered within 1.5 points, suggesting this benchmark is nearing saturation for frontier models.

34 models have been evaluated on AIME26. The benchmark falls in the Mathematics category. BenchAlign v6.0 gives AIME26 6% of the Math reference weight, so it moves the Math leaderboard and the overall ranking. Reference weights are relative weights in the calibrated model, not fixed shares of a score.

About AIME26

Year

2026

Tasks

Competition math problems

Format

Short-answer mathematics

Difficulty

Olympiad-style mathematics

AIME-style benchmarks remain one of the fastest ways to separate top reasoning models on olympiad-style math. AIME 2026 is a newer contest-year snapshot than the legacy AIME rows already tracked on BenchLM.

Freshness and provenance

Version

AIME26 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

Current

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does AIME26 measure?

A 2026 American Invitational Mathematics Examination snapshot used in frontier-model comparison tables for mathematical reasoning.

Which model scores highest on AIME26?

GLM-5.2 by Z.AI currently leads with a score of 99.2% on AIME26.

How many models are evaluated on AIME26?

34 AI models have published results on AIME26 in the BenchLM catalog.

Last updated: November 21, 2026 · BenchLM version AIME26 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.