Skip to main content
BenchLM

Vals MATH 500 (MATH 500)

We show this table for reference; we do not rank on it.

Data verified 36 confirmed releases in the last 30 daysFollow model changes

Academic math benchmark on probability, algebra, and trigonometry

MATH 500 score on MATH 500 — January 9, 2026

We mirror the published math 500 score view for MATH 500. Gemini 3 Pro Preview leads the public snapshot at 96.40%, followed by Grok 4 0709 (96.20%) and GPT-5 (96.00%). We do not use these results to rank models overall.

1

Gemini 3 Pro Preview

Google

96.40%
2

Grok 4 0709

xAI

96.20%
3

GPT-5

OpenAI

96.00%

60 modelsMathematicsCurrentDisplay onlyUpdated January 9, 2026

MATH 500 score table (60 models)

Score
1
Gemini 3 Pro PreviewGoogle
96.40%
2
Grok 4 0709xAI
96.20%
3
GPT-5OpenAI
96.00%
4
Claude Opus 4.1 20250805 ThinkingAnthropic
95.40%
5
Gemini 2.5 Pro Exp 03 25Google
95.20%
6
GPT-5 miniOpenAI · Closed
94.80%
7
GPT-OSS 120BOpenAI · Open weight
94.80%
8
o3OpenAI · Closed
94.60%
9
Qwen3 235b A22bFireworks
94.60%
10
O4 MiniOpenAI
94.20%
11
Grok 3 Mini Fast High ReasoningxAI
94.20%
12
GPT-OSS 20BOpenAI · Open weight
94.20%
13
Moonshotai Kimi K2 InstructTogether AI
94.20%
14
GLM-4.5Z.AI · Closed
94.00%
15
GPT-5 nanoOpenAI · Closed
93.80%
16
Claude Sonnet 4 20250514 ThinkingAnthropic
93.80%
17
Claude Opus 4.1Anthropic
93.00%
18
DeepSeek-R1DeepSeek · Open weight
92.20%
19
o3-miniOpenAI · Closed
91.80%
20
Gemini 2.5 Flash Preview 04 17 ThinkingGoogle
91.80%
21
Gemini 2.5 Flash Preview 04 17Google
91.60%
22
Claude 3.7 Sonnet 20250219 ThinkingAnthropic
91.60%
23
Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561 ThinkingTogether AI
91.40%
24
Claude Opus 4Anthropic
90.40%
25
o1OpenAI · Closed
90.40%
26
Claude Sonnet 4Anthropic
90.32%
27
Grok 3xAI
89.80%
28
Gemini 2.0 Flash ExpGoogle
89.00%
29
MiniMax M2.1MiniMax
89.00%
30
DeepSeek V3 0324Fireworks
88.60%
31
Gemini 2.0 Flash 001Google
88.00%
32
GPT-4.1 miniOpenAI · Closed
88.00%
33
GPT-4.1OpenAI · Closed
87.20%
34
Mistral Medium 2505Mistral
87.00%
35
Llama4 Maverick Instruct BasicFireworks
85.20%
36
Gemini 2.0 Flash Thinking Exp 01 21Google
84.60%
37
Gemini 1.5 Pro 002Google
82.80%
38
DeepSeek V3DeepSeek · Open weight
80.40%
39
GPT-4.1 nanoOpenAI · Closed
80.20%
40
Meta Llama Llama 4 Scout 17B 16E InstructTogether AI
79.20%
41
Gemini 1.5 Flash 002Google
78.80%
42
Grok 2 1212xAI
78.40%
43
Claude 3.7 SonnetAnthropic
76.80%
44
Command A 03 2025Cohere
76.20%
45
GPT-4oOpenAI · Closed
75.20%
46
Mistral Large 2411Mistral
74.40%
47
GPT-4oOpenAI · Closed
74.00%
48
Meta Llama Llama 3.3 70B Instruct TurboTogether AI
73.40%
49
GPT-4o miniOpenAI · Closed
72.60%
50
Claude 3.5 SonnetAnthropic · Closed
72.40%
51
Meta Llama Meta Llama 3.1 405B Instruct TurboTogether AI
71.40%
52
Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561Together AI
71.20%
53
Mistral Small 2402Mistral
70.60%
54
Grok 3 Mini Fast Low ReasoningxAI
70.20%
55
Mistral Small 2503Mistral
68.40%
56
Meta Llama Meta Llama 3.1 70B Instruct TurboTogether AI
65.00%
57
Claude 3.5 HaikuAnthropic
64.20%
58
Jamba Large 1.6AI21 Labs
54.80%
59
Meta Llama Meta Llama 3.1 8B Instruct TurboTogether AI
44.40%
60
Jamba Mini 1.6AI21 Labs
25.40%

How MATH 500 is shown here

BenchLM mirrors the public Vals AI MATH 500 leaderboard captured from https://www.vals.ai/benchmarks/math500 and updated by Vals on January 9, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

MATH 500 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.

Snapshot

60 Vals rows1 task viewspublic datasetTasks: OverallDisplay only
MATH 500 on Vals AIVals methodologyVals home

The published MATH 500 snapshot places Gemini 3 Pro Preview first at 96.40%. The third row is 0.40 points behind. The broader top-10 range is 2.20 points, so many of the published results sit in a relatively narrow band.

60 models have been evaluated on MATH 500. The benchmark falls in the Mathematics category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. MATH 500 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About MATH 500

Year

2026

Tasks

MATH 500 academic math problems

Format

Accuracy score

Difficulty

Advanced academic math

BenchLM mirrors the public Vals AI MATH 500 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.

MATH 500Public benchmark source

Freshness and provenance

Version

MATH 500 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.

Questions

What does MATH 500 measure?

Academic math benchmark on probability, algebra, and trigonometry

Which model leads the published MATH 500 snapshot?

Gemini 3 Pro Preview currently leads the published MATH 500 snapshot with 96.40% math 500 score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on MATH 500?

The January 9, 2026 snapshot contains 60 AI models.

Last updated: January 9, 2026 · mirrored from the public benchmark leaderboard

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 2,000+ readers.

One email each week. Unsubscribe anytime.