Gemma 3 27B

Name: Gemma 3 27B
Price: 0.1 USD
Author: Google

Google · Gemma 3 · released 2025-03-12 · data checked 2026-11-01 Needs verification

Gemma 3 27B is a 27.4B-parameter dense model from Google (United States), released 2025-03-12 under the Gemma license. It scores 32.7 on the Small LM Index, #32 of 34 medium models. It needs about 18.8 GB of memory at 4-bit with an 8K-token context and costs $0.10 per million tokens through the cheapest API provider.

Small LM Index

32.7

#32 of 34 scored medium models

Likely 25.7–40.6 · from 3 of 6 core benchmarks

Parameters

27.4B

Dense

Price per 1M tokens

$0.10

$0.080 in / $0.16 out · blended 3:1

Context window

128K

tokens

Benchmarks

Each bar shows this model's score. The tick marks the best medium model on that benchmark. Sources: 3 reported by Google. How we label sources.

MMLU-Pro Lab-reported Knowledge67.5

Class best: Qwen3.6 27B 86.2

GPQA Diamond Lab-reported Knowledge42.4

Class best: Nanbeige4.3 33B 89.5

LiveCodeBench Lab-reported Coding · LiveCodeBench v529.7

Class best: Qwen3.8 27B 90.3

Not reported: estimated

What the index model expects from this model's other scores. Not a measurement.

Humanity's Last Exam≈ 3

AIME≈ 39

SWE-bench Verified≈ 14

Specifications

Lab: Google (US · United States)
Architecture: Dense transformer
Total parameters: 27.4B
Active parameters: All (dense)
Size class: Medium
Context window: 128K tokens
Input modalities: text, image
Reasoning: Instruct (no thinking)
License: Gemma
Openness: Open weights
Memory to run: ~18.8 GB at 4-bit, 8K context (more sizes)
Hugging Face: 359.4K downloads (30 days) · 2.1K likes
API providers: 4 on OpenRouter
API price: $0.080 in / $0.16 out per 1M (OpenRouter, 2026-11-03)

Weights on Hugging Face ↗Try on OpenRouter ↗

Compare with

Gemma 3 27B vs Qwen3.8 27B (27B, index 77.0)
Gemma 3 27B vs Ternary Bonsai 2 27B (27.4B, index 74.6)
Gemma 3 27B vs Qwen3.6 27B (27B, index 74.0)
Gemma 3 27B vs Qwen3.5 27B (27B, index 70.6)

Sources

https://huggingface.co/google/gemma-3-27b-it

Something wrong here? Report it.

Run it locally

Uses google/gemma-3-27b-it-qat-q4_0-gguf (Q4_0). Swap the tag for another size: Q4_0.

Ollama

ollama run hf.co/google/gemma-3-27b-it-qat-q4_0-gguf:Q4_0

llama.cpp

llama-server -hf google/gemma-3-27b-it-qat-q4_0-gguf:Q4_0

Then open http://localhost:8080 for a chat window.

MLX (Mac)

pip install -U mlx-lm
mlx_lm.chat --model mlx-community/gemma-3-27b-it-4bit-DWQ

Text only; for images use mlx-vlm.

vLLM (GPU server)

vllm serve google/gemma-3-27b-it

Full-precision weights, for a server GPU.

Open in LM Studio

Memory needed: weights, context cache and runtime buffers.

Quantization	4K context	32K context	128K context
16-bit F16	57 GB	59 GB	67 GB
8-bit Q8_0	31 GB	33 GB	41 GB
4-bit Q4_K_M	18 GB	21 GB	29 GB

Check it against your GPU, Mac or phone →