Inference
EN
Google

Gemma 4 E2B

Gemma 4 E2B is a reasoning model from Google in the Gemma 4 family. 13 benchmarks count toward its index, in 7 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2026-04
RankedEnough results for an index.
IndexAll categories, out of 100.21.9±5.5Median 51.0
CoverageIndex weight that has results.90%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.—Median $0.60
Output / 1MUS dollars per 1M output tokens.—Median $2.55
ContextMaximum tokens in one request.128KMedian 262K
EloLMArena rating and rank.—Median 1422

Run locally

5.12B parameters
16-bit
11.8 GB
8-bit
6.7 GB
4-bit
4.3 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 8 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.2 counted25.3212 / 218
CodingCode writing and repair.1 counted20.5—
ReasoningLogic problems and puzzles.3 counted25.6216 / 227
MultimodalTasks with images and text.2 counted16.884 / 84
KnowledgeFacts and expert knowledge.3 counted22.1263 / 268
MultilingualTasks in many languages.1 counted5.0—
InstructionTasks with strict prompt rules.1 counted24.5—
MathMath problems.None counted——

Results

13 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Critical Physics TasksReasoning0.0%150 / 214
GDPval-AA normalizedAgentic0.0%101 / 133
Artificial Analysis IFBenchInstruction38.0%114 / 146
EuroEval FrenchMultilingual24.3% ±0.8144 / 178
Artificial Analysis Humanity's Last ExamKnowledge4.8%181 / 222
EuroEval PolishMultilingual20.2% ±0.9173 / 209
EuroEval SwedishMultilingual19.3% ±1.0162 / 195
EuroEval DutchMultilingual24.1% ±0.7146 / 175
EuroEval ItalianMultilingual16.6% ±0.9148 / 174
Massive Multitask Language Understanding ProfessionalKnowledge60.0%64 / 75
EuroEval PortugueseMultilingual19.5% ±0.8157 / 182
EuroEval SpanishMultilingual14.6% ±0.9152 / 176
τ²-Bench Tool-Agent-User EvaluationAgentic20.8%130 / 149
Artificial Analysis Long Context ReasoningReasoning16.3%195 / 217
EuroEval GermanMultilingual12.9% ±0.9173 / 191
Graduate-Level Google-Proof Q&AKnowledge43.4%129 / 141
Artificial Analysis GPQA DiamondKnowledge43.3%183 / 199
LiveCodeBench v6Coding44.0%55 / 59
Artificial Analysis MMMU-ProMultimodal44.6%107 / 113
ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard)Multimodal51.9%25 / 26
Massive Multi-discipline Multimodal Understanding ProMultimodal44.2%52 / 53
MRCRv2Reasoning19.1%14 / 14

Sources

BenchLMCC BY-NC 4.0EuroEvalMIT

History

5 releases
Index22.1 → 21.9
23 Oct 202612 Nov 2026

Questions

How good is Gemma 4 E2B?
Its index is 21.9 out of 100. That is place 277 of 283 ranked models.
What is Gemma 4 E2B best at?
Reasoning: 25.6 out of 100, place 216 of 227.
What is the context window of Gemma 4 E2B?
128K tokens.
Is Gemma 4 E2B open source?
Yes. Its weights are published.
When was Gemma 4 E2B released?
April 2026.

Badge

Gemma 4 E2B on Inference360: place 277 of 283, index 21.9
HTML
<a href="https://inference360.ai/models/gemma-4-e2b"><img src="https://inference360.ai/badges/gemma-4-e2b.svg" alt="Gemma 4 E2B on Inference360: place 277 of 283, index 21.9" height="20"></a>
Markdown
[![Gemma 4 E2B on Inference360: place 277 of 283, index 21.9](https://inference360.ai/badges/gemma-4-e2b.svg)](https://inference360.ai/models/gemma-4-e2b)

Compare with

Claude 3 Haiku22.0Phi-422.4Gemini 1.5 Flash20.2Llama 3.1 8B17.9Gemma 4 E4B28.2

More from Google

All models
Gemma 4 31B47.9Gemma 4 26B A4B44.5Gemma 4 E4B28.2Lyria 3 Pro PreviewUnrankedLyria 3 Clip PreviewUnranked