Model

Gemma 3 27B

OPEN

Google · Gemma 3 family · released Mar 12, 2025

Best-in-class single-GPU open VLM of early 2025; LMSYS Arena Elo 1338, top-10 among all models at launch.

ReasoningCodingVisionFunction callingTool useAgentic

1001.5

Elo · rank #212

Parameters

27.4B

Active params

27.4B (dense)

Context

128K tokens

Architecture

Dense decoder-only Transformer, GQA + 5:1 local:global attention, shared frozen SigLIP-400M vision encoder

License

Gemma

Languages

140+

API price (in/out)

No hosted API

Modalities

text · vision

Benchmark results

Bar shows position within the tracked field; marker = field best

Arena EloHuman preference1338#32

best: Claude Fable 5 · 1505

BIRD-SQLCoding54.4%#5

best: Command A · 59.5%

FACTS GroundingKnowledge74.9%#8

best: Gemini 2.5 Flash-Lite (Thinking) · 86.8%

GPQA DiamondReasoning42.4%#207

best: GPT-5.6 · 94.6%

LiveCodeBenchCoding29.7%#149

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MATH-500Math89.0%#51

best: GPT-5 · 99.4%

MMLU-ProKnowledge67.5%#100

best: Claude Fable 5 · 91.5%

MMLUKnowledge75.1%#125

best: OpenAI o3 · 92.9%

MMMUVision64.9%#66

best: Claude Fable 5 · 89.3%

SimpleQAKnowledge10.0%#37

best: GPT-4.5 · 62.5%

Run it locally

VRAM @ Q4

14.1 GB

VRAM @ FP16

54 GB

Fits on (Q4)

RTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Throughput data unavailable.

Quantizations

GGUF Q4_K_M · GGUF Q8_0 · AWQ · MLX · QAT Int4

Gemma 3 family

Elo progression across releases

Gemma 3 12BMar 2025944.3 Gemma 3 27BMar 20251001.5 Gemma 3 4BMar 2025680.3

Resources

HFgoogle/gemma-3-27b-it GHgoogle-deepmind/gemma DOCSai.google.dev/gemma/docs/core APIAvailable via Google API

API price weights · each benchmark row carries its own source badge (see methodology)