Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

DiffusionGemma 26B A4B

Google · 26B · A4B

Strongest in Knowledge work (#133 of 198, on 1 of its 2 boards), weakest in Agents (#190 of 289). Above par in 0 of 2 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and Claude Fable 5.1 and ahead of Ling 2.6 Flash and Qwen3 8B.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Agents46#190/289
Knowledge work42#133/198

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for DiffusionGemma 26B A4B, left for the other.

Gemma 4 31Bwon 2 · lost 0
Qwen3 30B A3B Thinkingwon 2 · lost 0
K2 Think V2won 2 · lost 0
DeepSeek V3won 2 · lost 0
Llama 3.1 8B Instructwon 2 · lost 0
K2 Horizon 0.9Bwon 2 · lost 0
Phi-4 Mini Instructwon 2 · lost 0
Nanbeige4.1 3Bwon 2 · lost 0
Ministral 3 8Bwon 2 · lost 0
Qwen3.5 0.8Bwon 2 · lost 0
Granite 4.2 3Bwon 2 · lost 0
MiniCPM-V 4.6 1.3Bwon 2 · lost 0
NVIDIA Nemotron 3 Nano 30B A3Bwon 2 · lost 0
Solar Pro 3won 2 · lost 0
Gemma 3 27B Itwon 2 · lost 0
O3 Miniwon 2 · lost 0
Ministral 3 14Bwon 2 · lost 0
Gemma 3 12B Itwon 2 · lost 0
DeepSeek R1won 2 · lost 0
Celeris 1won 2 · lost 0
Qwen3 235B A22B Thinkingwon 2 · lost 0
Mistral Small 3.2won 2 · lost 0
North Mini Codewon 2 · lost 0
Gemma 4 E2Bwon 2 · lost 0
Llama 4 Scout Instructwon 2 · lost 0
Granite 4.1 3Bwon 2 · lost 0
Trinity Largewon 2 · lost 0
Mistral Small 4won 2 · lost 0
Gemini 3.1 Flash Litewon 2 · lost 0
Mistral Large 3won 2 · lost 0
Llama 3.3 70B Instructwon 2 · lost 0
GPT-4.1 Nanowon 2 · lost 0
HyperNova 60B (High, Based on gpt-oss-120b)won 2 · lost 0
Nemotron 3 Nano 4Bwon 2 · lost 0
Mistral Small 3.1 24B Instructwon 2 · lost 0
Qwen3.5 2Bwon 2 · lost 0
Gemma 4 E4Bwon 2 · lost 0
GPT-4o Miniwon 2 · lost 0
Llama 4 Maverick Instructwon 2 · lost 0
GPT-4won 2 · lost 0
GPT-4owon 2 · lost 0
LFM2.5 2.6Bwon 2 · lost 0
Mercury 2won 2 · lost 0
Gemini 2.5 Prowon 2 · lost 0
Magistral Small 1.2won 2 · lost 0
Qwen3.5 9Bwon 2 · lost 0
Ministral 3 3Bwon 2 · lost 0
Granite 4.2 8Bwon 2 · lost 0
Mercury 2.5won 2 · lost 0
Command A+won 2 · lost 0
Gemma 4 26B A4Bwon 2 · lost 0
Qwen3 Next 80B A3B Thinkingwon 2 · lost 0
Qwen3 Coder Nextwon 2 · lost 0
Mistral Medium 3.1won 2 · lost 0
Magistral Medium 1.2won 2 · lost 0
Granite 4.2 30Bwon 2 · lost 0
Granite 4.1 30Bwon 2 · lost 0
Claude 3.5 Haikuwon 2 · lost 0
Nemotron 3 Nano Omni 30B A3B Reasoningwon 2 · lost 0
Nemotron Cascade 2 30B A3Bwon 2 · lost 0
Qwen3 14Bwon 2 · lost 0
Qwen3 8Bwon 2 · lost 0
Ling 2.6 Flashwon 2 · lost 0
NVIDIA Nemotron 3.5 Lightning 30B A3Bwon 1 · lost 1
NVIDIA Nemotron 3 Ultra 550B A55Bwon 1 · lost 1
Gemini 3.1 Prowon 1 · lost 1
Inklingwon 1 · lost 1
Mistral Large 4won 1 · lost 1
Nova 2 Litewon 1 · lost 1
Devstral Small 2won 1 · lost 1
Devstral 2won 1 · lost 1
Claude Haiku 4.5won 1 · lost 1
Ling 3.0 Tinywon 1 · lost 1
MiniCPM5 2Bwon 1 · lost 1
Qwen3 32Bwon 1 · lost 1
GPT-4.1 Miniwon 1 · lost 1
DeepSeek V3.1 Terminuswon 0 · lost 2
Nemotron 3.5 Super VLwon 0 · lost 2
GPT OSS 20Bwon 0 · lost 2
Beam 1.0won 0 · lost 2
K EXAONEwon 0 · lost 2
Muse Glimmerwon 0 · lost 2
Gemma 4 12Bwon 0 · lost 2
Mistral Medium 3.5won 0 · lost 2
Qwen3.5 122B A10Bwon 0 · lost 2
EXAONE 4.5 33Bwon 0 · lost 2
Nova 2.0 Pro Previewwon 0 · lost 2
KAT-Coder-Pro V1won 0 · lost 2
Qwen3.5 397B A17Bwon 0 · lost 2
NVIDIA Nemotron 3 Super 120B A12Bwon 0 · lost 2
Ring 2.6.1Twon 0 · lost 2
Qwen3.5 35B A3Bwon 0 · lost 2
GPT OSS 120Bwon 0 · lost 2
Mimo V2 Flashwon 0 · lost 2
Claude Sonnet 4won 0 · lost 2
Qwen3.6 35B A3Bwon 0 · lost 2
MiniMax M3won 0 · lost 2
GPT-5.4 Nanowon 0 · lost 2
LongCat 2.0won 0 · lost 2
Gemini 3.5 Flash Litewon 0 · lost 2
Muse Sparkwon 0 · lost 2
KAT Coder Pro V2won 0 · lost 2
GPT-5.5 Instantwon 0 · lost 2
GPT-5.3 Codexwon 0 · lost 2
K2 Horizon 3.7Bwon 0 · lost 2
A.X K2won 0 · lost 2
Grok 4.3won 0 · lost 2
MiniMax M2.7won 0 · lost 2
GPT-5.4 Miniwon 0 · lost 2
Mimo V2.5won 0 · lost 2
GPT-5 Miniwon 0 · lost 2
Qwen3.8 27Bwon 0 · lost 2
Ling 3.0 Flashwon 0 · lost 2
GPT-5.1won 0 · lost 2
Step 3.7 Flashwon 0 · lost 2
Qwen3.6 27Bwon 0 · lost 2
Muse Spark 1.2won 0 · lost 2
Gemini 3.5 Flashwon 0 · lost 2
GPT-5won 0 · lost 2
Kimi K2.7 Codewon 0 · lost 2
Nemotron 3.5 Bolt 74B-A9Bwon 0 · lost 2
Solar Open2 250Bwon 0 · lost 2
GPT-5.6 Lunawon 0 · lost 2
Qwen3.6 Pluswon 0 · lost 2
Solar Mini 4won 0 · lost 2
GLM-4.7won 0 · lost 2
GPT-5.5won 0 · lost 2
Mimo V2.5 Prowon 0 · lost 2
Muse Spark Openwon 0 · lost 2
Hy3won 0 · lost 2
Inkling Smallwon 0 · lost 2
GLM-5.3won 0 · lost 2
Qwen3.7 Maxwon 0 · lost 2
Grok Build 0.1won 0 · lost 2
Nex-N2-Pro (Based on Qwen3.5-397B-A17B)won 0 · lost 2
Step 5won 0 · lost 2
Grok 4.5won 0 · lost 2
K2 Horizon 7Bwon 0 · lost 2
GPT-6 Lunawon 0 · lost 2
Ling 3.0 Flash Finwon 0 · lost 2
GLM-5.1won 0 · lost 2
G9v3 39A5Bwon 0 · lost 2
Claude Sonnet 4.6won 0 · lost 2
Motif 3won 0 · lost 2
Gemini 3.7 Flashwon 0 · lost 2
Ling 3.0 Flash VLwon 0 · lost 2
Kimi K2.5won 0 · lost 2
Apodex 1.1won 0 · lost 2
Solar Pro4won 0 · lost 2
Gemini 3.6 Flashwon 0 · lost 2
DeepSeek V3.2won 0 · lost 2
Qwen3.7 Pluswon 0 · lost 2
GPT-5.4won 0 · lost 2
Quasar 438B (Max, Based on GLM-5.2)won 0 · lost 2
Kimi K3won 0 · lost 2
Kimi K2.6won 0 · lost 2
Agnes 2.5 Prowon 0 · lost 2
K2 Horizon MoVA 36B A4Bwon 0 · lost 2
Gemini 3.8 Flashwon 0 · lost 2
JT-4.1 Flash 236B A21Bwon 0 · lost 2
Qwen3.8 Maxwon 0 · lost 2
DeepSeek V4 Flash Visionwon 0 · lost 2
GPT-5.6 Terrawon 0 · lost 2
GLM-5.2won 0 · lost 2
Claude Opus 4.7won 0 · lost 2
GLM-5.3 Flashwon 0 · lost 2
GPT-5.6 Solwon 0 · lost 2
Claude Sonnet 4.5won 0 · lost 2
Muse Spark 1.1won 0 · lost 2
K2 Horizon 375B A23Bwon 0 · lost 2
DeepSeek V4 Flashwon 0 · lost 2
DeepSeek V4 Prowon 0 · lost 2
GLM-4.6won 0 · lost 2
Claude Sonnet 5won 0 · lost 2
Muse Spark 1.3won 0 · lost 2
MiniMax M3.1 Flashwon 0 · lost 2
Grok 4.6won 0 · lost 2
Ling 3.1 Flashwon 0 · lost 2
Qwen3.8 2.4T A95Bwon 0 · lost 2
Claude Opus 4.8won 0 · lost 2
Mimo V2.6 Flashwon 0 · lost 2
Mimo V2.6 Prowon 0 · lost 2
GPT-6 Solwon 0 · lost 2
DeepSeek V4.1 Flashwon 0 · lost 2
Grok 4.7won 0 · lost 2
Qwen3.8 Flash Nextwon 0 · lost 2
Claude Haiku 5.5won 0 · lost 2
Muse Wonderwon 0 · lost 2
Gemini 4 Argonwon 0 · lost 2
Muse Wonder Fullwon 0 · lost 2
GPT-6.1 Solwon 0 · lost 2
GPT-6 Astrawon 0 · lost 2
Claude Fable 5won 0 · lost 2
Claude Opus 5won 0 · lost 2
Claude Sonnet 5.5won 0 · lost 2
Claude Fable 5.1won 0 · lost 2
Claude Opus 5.5won 0 · lost 2

§ 3 · Sources

Where the numbers come from

1 publication, 1 figures. Every one links to the page it was read from.

GDPval-AA ↗1 measureread 2026-10-19
GDPval-AA 321

Badge

PublicAI Index badge for DiffusionGemma 26B A4B[![PublicAI Index](https://publicai.io/model-index/badge?model=diffusiongemma-26b-a4b)](https://publicai.io/model-index/m/diffusiongemma-26b-a4b)