[whatmodel.sh](../../README.md) / DiffusionGemma 26B-A4B # DiffusionGemma 26B-A4B Google · released 2026-06 · 25.82B params (3.8B active) · apache-2.0 **score 58.2** ±1.4 · comparison only: no current run option verified · from 12 benchmark results across 1 source · † no independently run results verified Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at _{● marks DiffusionGemma 26B-A4B among every model on that task or test, worst to best.}

task                vs the others          rank  counts  tests
agent tasks         ▁▁▂▆●▄▅▄▄▃▁▁   #215 of 354*    8.2%      1
coding              ▁▁▁▃▄▆█●▅▄▃▁   #227 of 539*    9.4%      1
math                ▁▁▃▆██●█▃▁▁▁   #217 of 530*     19%      2
vision & documents  ▁▁▂▃▂▃●▄▆██▄    #124 of 189     15%      3

agent tasks · #215 of 354 · 8.2% of its score

Beats [o1 (2024-12-17)](o1-2024-12-17.md), [o3-mini](o3-mini.md), [DeepSeek R1 0528](deepseek-r1-0528.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md), [Jamba 1.5 Mini](jamba-1-5-mini.md), [Mellum2 12B-A2.5B Instruct](mellum2-12b-a2-5b-instruct.md) here, though they rank lower overall.

test                        result  vs the others in its run  difficulty        counts  why
τ²-bench aggregate (maker)   56.2%  ▁▁▁·▁▁●▃▃▄█▁ beats 7/39   medium · avg 39%    8.2%  at its level · small run

coding · #227 of 539 · 9.4% of its score

Beats [Gemini 2.5 Flash](gemini-2-5-flash.md), [Qwen3 30B-A3B Instruct 2507](qwen3-30b-a3b-instruct-2507.md), [Qwen3.5 4B](qwen3-5-4b.md) here, though they rank higher overall. Trails [QwQ-32B-Preview](qwq-32b-preview.md), [Ministral 3 8B Reasoning 2512](ministral-3-8b-reasoning.md), [OpenThinker 32B](openthinker-32b.md) here, though they rank lower overall.

test                                result  vs the others in its run  difficulty        counts  why
LiveCodeBench v6 (window unspecif…   69.1%  ▂▃▄▄▅█●▆▄▇▄▆ beats 33/67  medium · avg 47%    9.4%  at its level · small run

math · #217 of 530 · 19% of its score

Beats [MiniMax-M2.1](minimax-m2-1.md), [DeepSeek V3 0324](deepseek-v3-0324.md), [DeepSeek V3 (Dec '24)](deepseek-v3.md) here, though they rank higher overall. Trails [Mixtral 8x7B v0.1 (base)](mixtral-8x7b-v0-1.md), [Phi-4-mini-reasoning 3.8B](phi-4-mini-reasoning.md), [OpenThinker3 7B](openthinker3-7b.md) here, though they rank lower overall.

test                 result  vs the others in its run  difficulty        counts  why
AIME 2026 (maker)     69.1%  ▁▁▁▂●▂▆█▆▂·▁ beats 8/52   medium · avg 51%     12%  sharp · at its level · small run
MATH-Vision (maker)   70.5%  ▁▁·▁··▄▁●▇▇█ beats 8/24   medium · avg 54%    7.0%  at its level · small run

vision & documents · #124 of 189 · 15% of its score

Beats [DeepSeek V4 Flash Vision (Exp)](deepseek-v4-flash-vision-exp.md) here, though they rank higher overall. Trails [Llama 4 Maverick (402B-A17B)](llama-4-maverick.md), [Qwen3-VL 8B Thinking](qwen3-vl-8b-thinking.md), [GPT-4o (2024-11-20)](gpt-4o-2024-11-20.md) here, though they rank lower overall.

test                        result  vs the others in its run  difficulty        counts  why
MATH-Vision (maker)          70.5%  ▁▁·▁··▄▁●▇▇█ beats 8/24   medium · avg 54%    7.0%  at its level · small run
MMMU Pro (maker)             54.3%  ▂▃▁▃▆●▃▇▇██▅ beats 17/52  medium · avg 46%    5.1%  at its level · small run
OmniDocBench 1.5 ↓ (maker)   31.9%  █●█·····████ beats 1/6    avg 36%             2.5%  blunt · small run

other tests (6)

test                            result  vs the others in its run    difficulty            counts  why
GPQA Diamond (maker)             73.2%  ▁▁▁▁▂▄▅▅●██▄ beats 168/322  hard · avg 49%           17%  sharp · at its level · small run
MedXPertQA MM (maker)            49.0%  ▂···▂·▃··●▂█ beats 6/13     hard · avg 33%           10%  sharp · at its level · small run
MMLU-Pro (maker)                 77.6%  ▁▁·▁▁▁▃▂▄●█▁ beats 111/188  medium · avg 59%        8.1%  at its level · small run
MRCR v2 8-needle, 128k (maker)   32.0%  █▄●█·▄···▄·▄ beats 3/8      hard · avg 22%          7.0%  at its level · small run
MMMLU (maker)                    81.5%  ▂▁▁▂▃▂▄▄▅●█▃ beats 33/46    medium · avg 67%        6.8%  at its level · small run
Humanity's Last Exam (maker)     11.0%  ▁▁▁▄●▆▇██▆▁▃ beats 23/111   very hard · avg 3.9%    5.8%  small run

## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its *counts* column above; here it is added up by who ran the tests. [Method](../method.md)

source                                    results  share of the score
google/diffusiongemma-26B-A4B-it model …       12   99% ▇▇▇▇▇▇▇▇▇▇▇▇
prior: pull toward the middle                        1% ▇
weighted average                                   = 58.2

_{Licences: google/diffusiongemma-26B-A4B-it model card ([Apache-2.0](https://huggingface.co/google/diffusiongemma-26B-A4B-it/blob/f7f5b7f5fa82ffc52addd066915886d497f5517b/README.md))} ## How to run it No verified API or local download yet. Model repository and documentation: [huggingface.co/google](https://huggingface.co/google/diffusiongemma-26B-A4B-it) [interactive version](https://whatmodel.sh/model/diffusiongemma-26b-a4b) · [JSON](https://whatmodel.sh/api?format=json&all=1)