Skip to main content
BenchLM

Vals MMMU (MMMU)

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

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

Multimodal Multi-task Benchmark

MMMU score on MMMU — September 1, 2026

We mirror the published mmmu score view for MMMU. Claude Fable 5.1 leads the public snapshot at 90.64%, followed by Claude Opus 5 (89.88%) and Claude Fable 5 (89.31%). We do not use these results to rank models overall.

93 modelsMultimodal & GroundedCurrentDisplay onlyUpdated September 1, 2026

MMMU score table (93 models)

Score
1
Claude Fable 5.1Anthropic · Closed
90.64%
2
Claude Opus 5Anthropic · Closed
89.88%
3
Claude Fable 5Anthropic · Closed
89.31%
4
Gemini 3.8 FlashGoogle · Closed
89.08%
5
Gemini 3.7 FlashGoogle · Closed
88.96%
6
GPT-5.6 SolOpenAI · Closed
88.84%
7
Gemini 3.6 FlashGoogle · Closed
88.38%
8
Gemini 3.5 FlashGoogle · Closed
88.27%
9
GPT-5.5OpenAI · Closed
88.27%
11
Kimi K3Moonshot AI · Closed
88.15%
12
Qwen3.8 MaxAlibaba · Open weight
88.03%
14
87.51%
15
GPT-5.4OpenAI · Closed
87.51%
16
Muse SparkMeta · Closed
87.40%
17
GPT-5.2OpenAI · Closed
86.67%
18
Muse Spark 1.1Meta · Closed
86.59%
19
Claude Opus 4.8Anthropic · Closed
86.59%
20
GPT-5.6 TerraOpenAI · Closed
86.47%
21
Kimi K2.6Moonshot AI · Open weight
86.30%
22
Muse Spark 1.2Meta · Closed
86.13%
23
GLM-5.3-FlashZ.AI · Open weight
86.01%
24
Claude Opus 4.7Anthropic · Closed
85.55%
25
GPT-5.6 LunaOpenAI · Closed
85.03%
26
84.33%
27
Qwen3.6 PlusAlibaba · Closed
84.16%
28
Qwen3.8-27BAlibaba · Open weight
83.93%
29
Claude Opus 4.6 (Adaptive)Anthropic · Closed
83.87%
30
Gemini 3.5 Flash-LiteGoogle · Closed
83.58%
31
Claude Sonnet 4.6Anthropic · Closed
83.58%
33
GPT-5.1OpenAI · Closed
83.18%
34
Grok 4.3xAI · Closed
83.06%
35
Claude Sonnet 5Anthropic · Closed
83.01%
36
Claude Opus 4.5 ThinkingAnthropic · Closed
82.95%
38
Qwen3.5 FlashAlibaba · Closed
81.91%
39
GPT-5OpenAI
81.50%
41
MiniMax M3MiniMax · Open weight
81.16%
42
Claude Opus 4.5Anthropic · Closed
81.10%
44
o3OpenAI · Closed
80.42%
45
MiMo-V2.5Xiaomi · Closed
80.00%
46
O4 MiniOpenAI
79.67%
48
Claude Sonnet 4.5 ThinkingAnthropic · Closed
79.31%
49
GPT-5.4 miniOpenAI · Closed
79.25%
50
GPT-5 miniOpenAI · Closed
78.91%
51
InklingThinking Machines Lab · Open weight
78.50%
53
o1OpenAI · Closed
77.41%
54
76.27%
58
73.72%
59
GPT-5.4 nanoOpenAI · Closed
73.58%
60
Claude Opus 4Anthropic
73.31%
61
72.78%
62
72.66%
64
72.39%
65
GPT-4.1OpenAI · Closed
72.39%
68
71.52%
69
GPT-5 nanoOpenAI · Closed
70.94%
70
GPT-4.1 miniOpenAI · Closed
70.54%
71
69.79%
72
Claude 3.5 SonnetAnthropic · Closed
68.80%
74
66.19%
75
65.51%
76
65.20%
77
64.57%
78
GPT-4oOpenAI · Closed
64.01%
81
62.97%
82
GPT-4oOpenAI · Closed
62.16%
83
Grok 4.5xAI · Closed
61.79%
84
60.08%
87
57.19%
88
GPT-4o miniOpenAI · Closed
56.56%
89
GPT-4.1 nanoOpenAI · Closed
55.05%
91
Claude Haiku 4.5 ThinkingAnthropic · Closed
46.07%
93
22.77%

How MMMU is shown here

BenchLM mirrors the public Vals AI MMMU leaderboard captured from https://www.vals.ai/benchmarks/mmmu and updated by Vals on September 1, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

MMMU 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

93 Vals rows1 task viewspublic datasetTasks: OverallDisplay only

The published MMMU snapshot places Claude Fable 5.1 first at 90.64%. The third row is 1.33 points behind. The broader top-10 range is 2.43 points, so many of the published results sit in a relatively narrow band.

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

About MMMU

Year

2026

Tasks

Multimodal academic task suite

Format

Accuracy score

Difficulty

Multimodal college-level reasoning

BenchLM mirrors the public Vals AI MMMU 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.

Freshness and provenance

Version

MMMU 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 MMMU measure?

Multimodal Multi-task Benchmark

Which model leads the published MMMU snapshot?

Claude Fable 5.1 currently leads the published MMMU snapshot with 90.64% mmmu score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on MMMU?

The September 1, 2026 snapshot contains 93 AI models.

Last updated: September 1, 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.