live
weekly refresh
basedagi.org
benchmark evidence

MMMU-Pro

MMMU-Pro: vision-only inputs prevent text shortcuts — harder multimodal reasoning across academic disciplines.

winner on MMMU-Pro
direct benchmark result, not a broad vertical composite | source row dated 2000-01-01
scored on 2000-01-01 · stale source data (9650d)
latest mapped results | top 20
#ModelScoreEvidenceTested
1OpenAI: GPT-5.5
Openai
83.2
model-only
independent_benchmark
2000-01-01
2OpenAI: GPT-5.4
Openai
81.2
model-only
independent_benchmark
2000-01-01
3OpenAI: GPT-5.2
Openai
79.5
model-only
independent_benchmark
2000-01-01
4OpenAI: GPT-5
Openai
78.4
model-only
independent_benchmark
2000-01-01
5Anthropic: Claude Opus 4.6
Anthropic
77.3
model-only
independent_benchmark
2000-01-01
6OpenAI: GPT-5.4 Mini
Openai
76.6
model-only
independent_benchmark
2000-01-01
7OpenAI: o3
Openai
76.4
model-only
independent_benchmark
2000-01-01
8Anthropic: Claude Sonnet 4.6
Anthropic
75.6
model-only
independent_benchmark
2000-01-01
9Meta: Llama 4 Maverick
Meta Llama
59.6
model-only
independent_benchmark
2000-01-01
what this result means

MMMU-Pro: vision-only inputs prevent text shortcuts — harder multimodal reasoning across academic disciplines.

This benchmark contributes direct public evidence. Read its scope before generalizing the result.

A win here is a win on MMMU-Pro. Broad task pages require independent corroboration before naming a general winner.

source record
category: multimodal
metric: accuracy
matched models: 9
latest source date: 2000-01-01
direction: higher is better
inspect upstream source ->