Skip to main content
BenchLM
Advertise

Benchmark profile

Vals AIME (AIME)

Challenging national math exam given to top high-school students

Data verified

How BenchLM shows AIME

BenchLM mirrors the public Vals AI AIME leaderboard captured from https://www.vals.ai/benchmarks/aime and updated by Vals on April 16, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

AIME is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.

96 Vals rows3 task viewspublic datasetTasks: Overall, AIME 2024, AIME 2025Display only
AIME on Vals AIVals methodologyVals home

AIME score on AIME — April 16, 2026

BenchLM mirrors the published aime score view for AIME. Gemini 3.1 Pro Preview leads the public snapshot at 98.13% , followed by GPT-5.2 (96.88%) and Muse Spark (96.88%). BenchLM does not use these results to rank models overall.

1

Gemini 3.1 Pro Preview

Google

google/gemini-3.1-pro-preview

98.13%
Overall —
2

GPT-5.2

OpenAI

openai/gpt-5.2-2025-12-11

96.88%
Overall —
3

Muse Spark

Meta

meta/muse_spark

96.88%
Overall —
96 modelsExternal benchmark mirrorsCurrentDisplay onlyUpdated April 16, 2026

AIME score table (96 models)

Score
1
Gemini 3.1 Pro PreviewGoogle
98.13%
2
GPT-5.2OpenAI
96.88%
3
Muse SparkMeta
96.88%
4
Gemini 3 Pro PreviewGoogle
96.68%
5
GPT-5.4OpenAI
96.67%
6
Grok 4.20 0309 ReasoningSpaceXAI
96.46%
7
Claude Opus 4.7Anthropic
96.25%
8
Gemini 3 Flash PreviewGoogle
95.63%
9
Claude Opus 4.6 ThinkingAnthropic
95.63%
10
GPT-5.4 MiniOpenAI
95.63%
11
Kimi K2.5 ThinkingMoonshot AI
95.63%
12
Claude Opus 4.5 20251101 ThinkingAnthropic
95.42%
13
Qwen3.6 PlusAlibaba
94.58%
14
GPT-5OpenAI
93.37%
15
GPT-5.1OpenAI
93.33%
16
GLM 4.7Zhipu AI
93.33%
17
GLM 4.6Zhipu AI
92.71%
18
GPT Oss 120bFireworks AI
92.60%
19
Qwen3.5 FlashAlibaba
92.50%
20
Claude Sonnet 4.6Anthropic
92.29%
21
Grok 4.1 Fast ReasoningSpaceXAI
91.88%
22
GLM 5.1Zhipu AI
91.88%
23
GLM 5 ThinkingZhipu AI
91.67%
24
GPT-5 MiniOpenAI
91.46%
25
Grok 4 Fast ReasoningSpaceXAI
91.25%
26
MiniMax M2.7MiniMax
91.04%
27
Grok 4 0709SpaceXAI
90.56%
28
GPT-5.4 NanoOpenAI
88.75%
29
MiniMax M2.5MiniMax
88.75%
30
Claude Sonnet 4.5 20250929 ThinkingAnthropic
88.19%
31
GLM 4.5Zhipu AI
86.67%
32
O3 MiniOpenAI
86.46%
33
Qwen3.5 Plus ThinkingAlibaba
86.04%
34
GPT Oss 20bFireworks AI
86.04%
35
Gemini 2.5 Pro Exp 03 25Google
85.83%
36
Kimi K2 ThinkingMoonshot AI
85.42%
37
O3OpenAI
85.28%
38
Grok 3 Mini Fast High ReasoningSpaceXAI
85.00%
39
DeepSeek V3p2 ThinkingFireworks AI
84.58%
40
Qwen3 235b A22bFireworks AI
83.96%
41
O4 MiniOpenAI
83.67%
42
Magistral Medium 2509Mistral AI
83.54%
43
Gemini 3.1 Flash Lite PreviewGoogle
83.33%
44
Claude Haiku 4.5 20251001 ThinkingAnthropic
82.71%
45
GPT-5 NanoOpenAI
81.18%
46
Qwen3 MaxAlibaba
81.04%
47
Magistral Small 2509Mistral AI
80.68%
48
Claude Opus 4.1 20250805 ThinkingAnthropic
78.18%
49
MiniMax M2.1MiniMax
77.92%
50
Claude Opus 4.5Anthropic
76.88%
51
Claude Sonnet 4 20250514 ThinkingAnthropic
76.25%
52
DeepSeek R1Fireworks AI
73.96%
53
O1OpenAI
71.46%
54
Grok 3 Mini Fast Low ReasoningSpaceXAI
70.63%
55
DeepSeek V3p2Fireworks AI
64.79%
56
Moonshotai Kimi K2 InstructTogether AI
62.71%
57
Qwen3 Max PreviewAlibaba
60.69%
58
Grok 3SpaceXAI
58.75%
59
Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561 ThinkingTogether AI
53.54%
60
DeepSeek V3 0324Fireworks AI
52.20%
61
Gemini 2.5 Flash Preview 09 2025 ThinkingGoogle
51.46%
62
Gemini 2.5 Flash Preview 09 2025Google
49.79%
63
GPT-4.1 MiniOpenAI
49.38%
64
Claude 3.7 Sonnet 20250219 ThinkingAnthropic
44.58%
65
Claude Opus 4.1Anthropic
44.24%
66
Mistral Large 2512Mistral AI
42.92%
67
Mistral Medium 2505Mistral AI
42.29%
68
Gemini 2.5 Flash Lite Preview 09 2025 ThinkingGoogle
42.08%
69
Claude Opus 4Anthropic
41.25%
70
GPT-4.1OpenAI
39.58%
71
Claude Sonnet 4Anthropic
38.54%
72
Grok 4 Fast Non ReasoningSpaceXAI
33.33%
73
Gemini 2.0 Flash 001Google
29.79%
74
DeepSeek V3Fireworks AI
27.50%
75
Grok 4.1 Fast Non ReasoningSpaceXAI
27.08%
76
GPT-4.1 NanoOpenAI
26.46%
77
Gemini 2.5 Flash Lite Preview 09 2025Google
26.25%
78
Llama4 Maverick Instruct BasicFireworks AI
25.21%
79
Claude 3.7 SonnetAnthropic
22.29%
80
Meta Llama Llama 4 Scout 17B 16E InstructTogether AI
18.96%
81
Gemini 1.5 Pro 002Google
18.75%
82
Gemini 1.5 Flash 002Google
17.29%
83
Meta Llama Llama 3.3 70B Instruct TurboTogether AI
16.04%
84
Grok 2 1212SpaceXAI
15.21%
85
GPT-4oOpenAI
13.96%
86
Command A 03 2025Cohere
13.33%
87
GPT-4oOpenAI
11.88%
88
GPT-4o MiniOpenAI
11.46%
89
Claude 3.5 SonnetAnthropic
10.00%
90
Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561Together AI
9.38%
91
Mistral Large 2411Mistral AI
9.17%
92
Mistral Small 2402Mistral AI
5.63%
93
Mistral Small 2503Mistral AI
3.54%
94
Claude 3.5 HaikuAnthropic
3.33%
95
Jamba Mini 1.6AI21 Labs
0.42%
96
Jamba Large 1.6AI21 Labs
0.42%

The published AIME snapshot places Gemini 3.1 Pro Preview first at 98.13%. The third row is 1.25 points behind. The broader top-10 range is 2.50 points, so many of the published results sit in a relatively narrow band.

96 models have been evaluated on AIME. The benchmark falls in the External benchmark mirrors category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. AIME is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About AIME

Year

2026

Tasks

AIME math problems

Format

Accuracy score

Difficulty

Competition math

BenchLM mirrors the public Vals AI AIME leaderboard as display-only external evidence. The captured snapshot preserves overall scores, task-level scores where Vals publishes them, uncertainty, latency, and cost-per-test metadata. It is excluded from BenchLM weighted rankings.

AIMEPublic benchmark source

BenchLM freshness & provenance

Version

AIME 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

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.

FAQ

What does AIME measure?

Challenging national math exam given to top high-school students

Which model leads the published AIME snapshot?

Gemini 3.1 Pro Preview currently leads the published AIME snapshot with 98.13% aime score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on AIME?

96 AI models are included in BenchLM's mirrored AIME snapshot, based on the public leaderboard captured on April 16, 2026.

Last updated: April 16, 2026 · mirrored from the public benchmark leaderboard

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.