Benchmarks / Multimodal
MMMU
Massive multi-discipline multimodal understanding benchmark.
maintainer unknown · 82 observations on record
Versions
1 version
| Version | Tasks |
|---|---|
| MMMU current mmmu | — |
Current leaderboard
MMMU · Accuracy · top 25 of 82 systems
| # | System | Accuracy % |
|---|---|---|
| 1 | Claude Fable 5max effort Anthropic | 89.3% |
| 2 | GPT 5.6 Solmax effort OpenAI | 88.8% |
| 3 | Gemini 3.5 Flashhigh effort Google DeepMind | 88.3% |
| 4 | GPT 5.5 (2026-04-22)xhigh effort OpenAI | 88.3% |
| 5 | Gemini 3.1 Pro Previewhigh effort Google DeepMind | 88.2% |
| 6 | Gemini 3 Flash Previewhigh effort Google DeepMind | 87.6% |
| 7 | GPT 5.4 (2026-03-05)xhigh effort OpenAI | 87.5% |
| 8 | Gemini 3 Pro Previewhigh effort Google DeepMind | 87.5% |
| 9 | Muse Spark Meta AI | 87.4% |
| 10 | GPT 5.2 (2025-12-11)xhigh effort OpenAI | 86.7% |
| 11 | Claude Opus 4.8max effort Anthropic | 86.6% |
| 12 | Muse Spark 1.1xhigh effort Meta AI | 86.6% |
| 13 | GPT 5.6 Terraxhigh effort OpenAI | 86.5% |
| 14 | Kimi K2 6 Moonshot AI | 86.3% |
| 15 | Claude Opus 4.7max effort Anthropic | 85.5% |
| 16 | GPT 5.6 Lunamax effort OpenAI | 85.0% |
| 17 | Kimi K2 5thinking Moonshot AI | 84.3% |
| 18 | Qwen3 6 Plus Alibaba | 84.2% |
| 19 | Claude Opus 4.6max effort Anthropic | 83.9% |
| 20 | Claude Sonnet 4.6max effort Anthropic | 83.6% |
| 21 | Grok 4.20 0309 Reasoning xAI | 83.5% |
| 22 | GPT 5.1 (2025-11-13)high effort OpenAI | 83.2% |
| 23 | Grok 4.3 xAI | 83.1% |
| 24 | Claude Sonnet 5max effort Anthropic | 83.0% |
| 25 | Claude Opus 4.5 20251101high effort Anthropic | 82.9% |
One row per evaluated system — reasoning-effort variants rank separately.