2026-09-10MODELDeepSeek-V4.1-Flash released with extreme agent memory efficiency and evaluated across VALS • 2026-09-03MODELGPT-6 Astra released by OpenAI - 71.7% on EMB, 63.3% on Tax Agent Bench • 2026-09-02MODELGoogle Gemini 3.8 Flash takes #1 rank on Finance Agent (v2) with 61.4% and 72.2% on EMB • 2026-09-02MODELMeta releases Muse Spark 1.3 & 1.3 Max - 59.0%-60.0% on Finance Agent (v2) • 2026-09-01MODELClaude Fable 5.1 leads Tax Agent Bench (77.6%) and EMB (76.7%) with 75% cheaper cache reads • 2026-09-11BENCHTax Agent Bench and FAB v2 benchmarks updated with latest frontier model rankings • 2026-09-15UPDATEPlatform updated: 43 models tracked across 41 independent benchmarks • 2026-09-10MODELDeepSeek-V4.1-Flash released with extreme agent memory efficiency and evaluated across VALS • 2026-09-03MODELGPT-6 Astra released by OpenAI - 71.7% on EMB, 63.3% on Tax Agent Bench • 2026-09-02MODELGoogle Gemini 3.8 Flash takes #1 rank on Finance Agent (v2) with 61.4% and 72.2% on EMB • 2026-09-02MODELMeta releases Muse Spark 1.3 & 1.3 Max - 59.0%-60.0% on Finance Agent (v2) • 2026-09-01MODELClaude Fable 5.1 leads Tax Agent Bench (77.6%) and EMB (76.7%) with 75% cheaper cache reads • 2026-09-11BENCHTax Agent Bench and FAB v2 benchmarks updated with latest frontier model rankings • 2026-09-15UPDATEPlatform updated: 43 models tracked across 41 independent benchmarks •
Home/Benchmarks/MMMU
Tier 3documentQAnumerical

MMMU

Massive Multi-discipline Multimodal Understanding and Reasoning benchmark for expert AGI.

11,500 tasks
14 models evaluated
Updated 2026
arXiv:2311.16502
View Source →Submit your scoreMethodology
89.9%
Top Score
Top model
Claude Opus 5
Model Rankings
financebenchmark.ai · MMMU
Anthropic
Anthropic
OpenAI
Moonshot AI
Anthropic
Google
Google
OpenAI
OpenAI
xAI
Thinking Machines
Meta
Meta
Alibaba
83.6%
🥇AnthropicAnthropicClaude Opus 589.9%
🥈AnthropicAnthropicClaude Sonnet 583.0%
🥉OpenAIOpenAIGPT-5.6 Sol88.8%
4Moonshot AIMoonshot AIKimi K3OPEN88.2%
5AnthropicAnthropicClaude Fable 589.3%
6GoogleGoogleGemini 3.5 Flash Lite83.6%
7GoogleGoogleGemini 3.6 Flash86.7%
8OpenAIOpenAIGPT-5.6 Luna85.0%
9OpenAIOpenAIGPT-5.6 Terra86.5%
10xAIxAIGrok 4.561.8%
11Thinking MachinesThinking MachinesInklingOPEN78.5%
12MetaMetaMuse Spark 1.186.6%
13MetaMetaMuse Spark 1.286.1%
14AlibabaAlibabaQwen 3.8 MaxOPEN88.0%
Full Results Table
#ModelOverallSourceDate
🥇
AnthropicClaude Opus 5
Anthropic
89.9arXiv:2311.165022026-07
🥈
AnthropicClaude Fable 5
Anthropic
89.3arXiv:2311.165022026-06
🥉
OpenAIGPT-5.6 Sol
OpenAI
88.8arXiv:2311.165022026-07
4
Moonshot AIKimi K3OPEN
Moonshot AI
88.2arXiv:2311.165022026-07
5
AlibabaQwen 3.8 MaxOPEN
Alibaba
88.0arXiv:2311.165022026-08
6
GoogleGemini 3.6 Flash
Google
86.7arXiv:2311.165022026-07
7
MetaMuse Spark 1.1
Meta
86.6arXiv:2311.165022026-07
8
OpenAIGPT-5.6 Terra
OpenAI
86.5arXiv:2311.165022026-07
9
MetaMuse Spark 1.2
Meta
86.1arXiv:2311.165022026-08
10
OpenAIGPT-5.6 Luna
OpenAI
85.0arXiv:2311.165022026-07
11
GoogleGemini 3.5 Flash Lite
Google
83.6arXiv:2311.165022026-07
12
AnthropicClaude Sonnet 5
Anthropic
83.0arXiv:2311.165022026-06
13
Thinking MachinesInklingOPEN
Thinking Machines
78.5arXiv:2311.165022026-07
14
xAIGrok 4.5
xAI
61.8arXiv:2311.165022026-07
All evaluated models
Anthropic
Claude Opus 5
89.9%
Anthropic
Claude Sonnet 5
83.0%
OpenAI
GPT-5.6 Sol
88.8%
Moonshot AI
Kimi K3
88.2%
Anthropic
Claude Fable 5
89.3%
Google
Gemini 3.5 Flash Lite
83.6%
Google
Gemini 3.6 Flash
86.7%
OpenAI
GPT-5.6 Luna
85.0%
OpenAI
GPT-5.6 Terra
86.5%
xAI
Grok 4.5
61.8%
Thinking Machines
Inkling
78.5%
Meta
Muse Spark 1.1
86.6%
Meta
Muse Spark 1.2
86.1%
Alibaba
Qwen 3.8 Max
88.0%