Inference
EN
Google

Gemma 3 27B

Gemma 3 27B is a standard model from Google. 13 benchmarks count toward its index, in 8 categories.

  • Open weights
  • Standard
  • Input: text
  • Released 2025-03
RankedEnough results for an index.
IndexAll categories, out of 100.25.3±4.5Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.31/sMedian 67/s
First tokenSeconds to the first output token.2.0 sMedian 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.32KMedian 262K
EloLMArena rating and rank.—Median 1422

Run locally

27.43B parameters
16-bit
58.6 GB
8-bit
31.6 GB
4-bit
18.5 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 24 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.2 counted21.9218 / 218
CodingCode writing and repair.1 counted26.2—
ReasoningLogic problems and puzzles.2 counted24.6217 / 227
MultimodalTasks with images and text.1 counted17.6—
KnowledgeFacts and expert knowledge.3 counted24.9257 / 268
MultilingualTasks in many languages.1 counted43.7—
InstructionTasks with strict prompt rules.1 counted18.7—
MathMath problems.2 counted28.5147 / 168

Results

13 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
EuroEval SwedishMultilingualAbove fitFar above this model's index.58.0% ±0.241 / 195
EuroEval GermanMultilingual47.2% ±0.549 / 191
EuroEval SpanishMultilingual49.2% ±0.447 / 176
EuroEval PolishMultilingualAbove fitFar above this model's index.51.8% ±0.361 / 209
EuroEval FrenchMultilingualAbove fitFar above this model's index.57.5% ±0.353 / 178
EuroEval PortugueseMultilingualAbove fitFar above this model's index.53.5% ±0.457 / 182
MATH level 5Math74.0% ±1.131 / 88
GPQA diamond · Epoch AIKnowledge47.7% ±2.7152 / 217
Critical Physics TasksReasoning0.0%150 / 214
OTIS Mock AIME 2024-2025Math22.5% ±5.3143 / 200
GDPval-AA normalizedAgentic0.0%101 / 133
Artificial Analysis Humanity's Last ExamKnowledge4.4%189 / 222
Artificial Analysis IFBenchInstruction31.8%130 / 146
Artificial Analysis Long Context ReasoningReasoning7.3%200 / 217
Artificial Analysis GPQA DiamondKnowledge42.8%184 / 199
Artificial Analysis MMMU-ProMultimodal48.0%105 / 113
Artificial Analysis Omniscience AccuracyKnowledge13.0%158 / 164
τ²-Bench Tool-Agent-User EvaluationAgentic10.5%145 / 149
Artificial Analysis SciCodeCoding23.3%111 / 114

Sources

BenchLMCC BY-NC 4.0Epoch AICC BYEuroEvalMIT

History

5 releases
Index26.0 → 25.3
23 Oct 202612 Nov 2026

Questions

How good is Gemma 3 27B?
Its index is 25.3 out of 100. That is place 268 of 283 ranked models.
What is Gemma 3 27B best at?
Math: 28.5 out of 100, place 147 of 168.
How fast is Gemma 3 27B?
31 output tokens per second, 2.0 s to the first token.
What is the context window of Gemma 3 27B?
32K tokens.
Is Gemma 3 27B open source?
Yes. Its weights are published.
When was Gemma 3 27B released?
March 2025.

Badge

Gemma 3 27B on Inference360: place 268 of 283, index 25.3
HTML
<a href="https://inference360.ai/models/gemma-3-27b"><img src="https://inference360.ai/badges/gemma-3-27b.svg" alt="Gemma 3 27B on Inference360: place 268 of 283, index 25.3" height="20"></a>
Markdown
[![Gemma 3 27B on Inference360: place 268 of 283, index 25.3](https://inference360.ai/badges/gemma-3-27b.svg)](https://inference360.ai/models/gemma-3-27b)

Compare with

Nova Pro25.6Llama 3.1 70B24.8GPT-4o mini26.2Gemini 2.5 Flash35.0Gemini 2.5 Pro46.6

More from Google

All models
Gemini 2.5 Pro46.6Gemini 2.0 Flash-LiteUnrankedGemini 2.0 Flash29.5Gemini 2.5 Pro Preview 06-05UnrankedGemini 2.5 Flash35.0