DiffusionGemma 26B A4B

DiffusionGemma 26B A4B ranks first on 2 of the 22 benchmarks it has scores on.

Google DeepMind Released 10 Jun 2026 Family Gemma 4 The lab's model card

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Google DeepMind Jun 2026

Open weights Mid-size

No Epoch Capabilities Index score yet

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Benchmark	Score	Rank	Source
IFEval	97.4%	#3 of 126	Model card
MGSM	84.8%	#25 of 52	Model card
BIG-Bench Extra Hard	47.6%	#6 of 11	Model card
MMLU-Pro	77.6%	#89 of 151	Model card
MMMLU	81.5%	#37 of 62	Model card
Humanity's Last Exam	11.9%	#126 of 147	Model card
MRCR v2	32%	#25 of 27	Model card

Benchmark	Score	Rank	Source
GSM8K	96.3%	#5 of 87	Model card
HiddenMath	80.6%	#1 of 11	Model card
GPQA Diamond	73.2%	#151 of 284	Model card
PutnamBench	67.4%	#4 of 7	Model card
AIME 2026	69.1%	#33 of 34	Model card

Benchmark	Score	Rank	Source
HumanEval	94.5%	#5 of 95	Model card
Natural2Code	94%	#1 of 9	Model card
LBPP (Less Basic Python Problems)	81%	#1 of 2	Model card
BigCodeBench	46%	#6 of 11	Model card
LiveCodeBench v6	69.1%	#50 of 76	Model card
Codeforces rating	1,429	#38 of 41	Model card

Benchmark	Score		Rank	Source
tau2-bench	56.2%		#63 of 69	Model card

Benchmark	Score		Rank	Source
MedXpertQA-MM	49%		#10 of 12	Model card

Benchmark	Score	Rank	Source
OmniDocBench v1.5	0.32	#6 of 13	Model card
MATH-Vision (MathVision)	70.5%	#24 of 42	Model card
MMMU-Pro	54.3%	#62 of 72	Model card

The benchmarks the lab chose to headline, with the scores it gave.

Every result in the longer documents the lab published with it. 9 of them aren't in the launch post.