The LLM benchmark aggregator.

Gemma 4 E4B

Google

Strongest in Secure code (#29 of 274), weakest in Fairness (#280 of 300). Above par in 2 of 13 scopes. Among the models it meets almost everywhere, it finishes behind Grok 4.7 and Gemini 3 Pro and ahead of MiniMax M3 and Gemma 3 27B It.

§ 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.

Safety48.3−13.2#228/3371/3

Secure code63.5−3#29/2741/1

Toxicity avoidance56.1−2.9#75/2721/1

Harm refusal46.8−14.8#212/3001/2

Jailbreak resistance38.4−26.5#226/2721/1

Fairness39.7−30.3#280/3001/2

Agents41.2−26.9#230/2681/5

Knowledge work36.9−36.7#149/1781/2

Tool use37.8−36.2—/81✱0/1

Reasoning47.5−19.4—/178✱0/4

Science42.2−21.3—/122✱0/1

Coding45.6−23.5—/165✱0/5

Agentic coding44.4−24.3—/157✱0/4

§ 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 Gemma 4 E4B, left for the other.

Jamba 1.6 Miniwon 6 · lost 2

Command R Pluswon 6 · lost 2

Grok 3won 6 · lost 2

QwQ 32Bwon 6 · lost 2

DeepSeek V3won 9 · lost 3

Command Awon 9 · lost 3

DeepSeek R1won 8 · lost 3

Grok Build 0.1won 8 · lost 3

Llama 3.1 8B Instructwon 6 · lost 3

Qwen3 8Bwon 6 · lost 3

Nemotron 3 Ultra 550B A55Bwon 7 · lost 4

Jamba 1.6 Largewon 5 · lost 3

Phi-4 Mini Instructwon 5 · lost 3

Gemma 4 E2Bwon 5 · lost 3

Command R7Bwon 5 · lost 3

Qwen3 30B A3B Instructwon 5 · lost 3

Ling 3.0 Flash VLwon 5 · lost 3

GPT-4.1 Nanowon 8 · lost 5

Ling 3.0 Flashwon 7 · lost 5

Grok 4.1 Fastwon 7 · lost 5

Mimo V2.6 Flashwon 5 · lost 4

Gemma 3 27B Itwon 6 · lost 5

MiniMax M3won 7 · lost 6

Gemini 1.5 Flashwon 4 · lost 4

Llama 3.3 Instruct Turbo (70B)won 4 · lost 4

Gemini 1.5 Prowon 4 · lost 4

GLM-4.5won 4 · lost 4

Llama 3.2 1B Instructwon 4 · lost 4

NVIDIA Nemotron 3 Super 120B A12Bwon 4 · lost 4

GPT-4owon 5 · lost 5

Gemini 2.0 Flashwon 5 · lost 5

Mistral Smallwon 5 · lost 5

Mistral Largewon 5 · lost 5

Grok 4.3won 6 · lost 6

Llama 4 Maverick Instructwon 6 · lost 7

GLM-4.6won 6 · lost 7

Llama 4 Scout Instructwon 5 · lost 6

O3 Miniwon 5 · lost 7

GPT-4o Miniwon 5 · lost 7

Gemini 3.5 Flashwon 5 · lost 7

Qwen3.7 Maxwon 5 · lost 7

Gemini 3.6 Flashwon 5 · lost 7

Kimi K2.6won 5 · lost 7

Grok 4won 5 · lost 7

GLM-5.3 Flashwon 5 · lost 7

Kimi K2 Thinkingwon 4 · lost 6

GLM-5.2won 5 · lost 8

GPT-4 Turbowon 3 · lost 5

Qwen Maxwon 3 · lost 5

Gemma 3 4B Itwon 3 · lost 5

Kimi K2.7 Codewon 4 · lost 7

O1 Miniwon 3 · lost 6

Gemma 3 12B Itwon 3 · lost 6

Mimo V2.5 Prowon 4 · lost 8

Gemini 2.5 Flashwon 4 · lost 8

Kimi K2won 4 · lost 8

DeepSeek V4 Flashwon 4 · lost 8

DeepSeek V3.2won 4 · lost 9

Claude 3.5 Sonnetwon 3 · lost 7

GPT OSS 20Bwon 3 · lost 7

GPT-5 Nanowon 3 · lost 7

Qwen3 235B A22B Instructwon 3 · lost 8

GPT-4won 2 · lost 6

Llama 3.2 3B Instructwon 2 · lost 6

Nova Microwon 2 · lost 6

Nova Prowon 2 · lost 6

Granite 3.2 8B Instructwon 2 · lost 6

Qwen3.5 122B A10Bwon 2 · lost 6

Gemma 4 26B A4Bwon 2 · lost 6

Hy3won 2 · lost 6

Gemini 2.5 Prowon 3 · lost 9

Claude 3.5 Haikuwon 3 · lost 9

GPT OSS 120Bwon 3 · lost 9

Kimi K2.5won 3 · lost 9

GPT-4.1won 3 · lost 9

GPT-4.1 Miniwon 3 · lost 10

GPT-5 Miniwon 3 · lost 10

Claude Haiku 4.5won 3 · lost 10

O1won 2 · lost 8

Claude Opus 4won 2 · lost 8

Gemma 4 12Bwon 2 · lost 9

Claude 3.7 Sonnetwon 2 · lost 9

GPT-6 Solwon 2 · lost 9

Mimo V2.6 Prowon 2 · lost 9

MiniMax M2.7won 2 · lost 10

O3won 2 · lost 10

GLM-5.1won 2 · lost 10

O4 Miniwon 2 · lost 10

Muse Spark 1.1won 2 · lost 10

Gemini 3.8 Flashwon 2 · lost 10

DeepSeek V4 Prowon 2 · lost 10

GPT-5.6 Solwon 2 · lost 10

Gemma 4 31Bwon 2 · lost 11

Inklingwon 2 · lost 11

GPT-5.6 Lunawon 2 · lost 11

Claude Opus 4.6won 1 · lost 7

Granite 3.1 8B Instructwon 1 · lost 7

Phi 4won 1 · lost 7

Qwen3.5 397B A17Bwon 1 · lost 7

Qwen3.5 35B A3Bwon 1 · lost 7

Claude Sonnet 4.6won 1 · lost 9

GPT-6 Lunawon 1 · lost 10

Claude Opus 5.5won 1 · lost 10

Claude Sonnet 4won 1 · lost 11

GPT-5.1won 1 · lost 11

GPT-5won 1 · lost 11

Grok 4.5won 1 · lost 11

GPT-5.4won 1 · lost 11

GPT-5.2won 1 · lost 11

GPT-5.5won 1 · lost 11

Claude Opus 4.7won 1 · lost 11

Claude Opus 4.8won 1 · lost 11

Grok 4.6won 1 · lost 11

Qwen3.8 Maxwon 1 · lost 11

Claude Opus 5won 1 · lost 11

Claude Sonnet 4.5won 1 · lost 12

GPT-5.6 Terrawon 1 · lost 12

Claude Sonnet 5won 1 · lost 12

Gemini 3 Prowon 0 · lost 9

Grok 4.7won 0 · lost 11

§ 3 · Sources

Where the numbers come from

3 publications, 10 figures. Every one links to the page it was read from.

GDPval-AA ↗1 measureread 2026-09-26

GDPval-AA -51

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26

Enkrypt · Jailbreak risk 22.6%Enkrypt · Harmful content risk 14.4%Enkrypt · CBRN risk 33.2%Enkrypt · Toxicity risk 2%Enkrypt · Bias risk 93.8%Enkrypt · Insecure code risk 6.2%

✱✱ IFM — K2 Horizon launch ↗the model’s own publisher3 measuresread 2026-09-26

tau3-Banking 5%Terminal-Bench 2.1 2%GPQA Diamond 52%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=gemma-4-e4b)](https://publicai.io/model-index/m/gemma-4-e4b)