Google DeepMind

Gemma 3 27B

Tool callingStructured outputImage input
Gemma 3 27B launch post cover

Largest open Gemma 3 instruction model for multilingual text generation and visual understanding

Context
131K
Released
1y ago
Parameters
27B
Input / 1M tokens
$0.08
Output / 1M tokens
$0.45
Cached input / 1M tokens
$0.04

Sources:models.devOpenRouter

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.

Other published benchmarks

EvaluationReported score and source
GPQA diamond
47.7%
Epoch AI · gemma-3-27b-it
Epoch AI evaluation
OTIS Mock AIME 2024-2025
22.5%
Epoch AI · gemma-3-27b-it
Epoch AI evaluation

Benchmark compilation: Epoch AI, Capabilities & benchmarking (CC BY attribution). Scores are converted to percentages and model names matched to the Vertical Tabs catalog. Individual evaluations are credited to their original sources. External results may carry separate licences and evaluation configurations; do not interpret scores from different harnesses as directly comparable.

Specifications

Model family
Gemma
Type
Language model
Predecessor
—
Successor
—
Released
March 12, 2025
Model ID
google/gemma-3-27b-it
Openness
Open weights
License
—
Parameters
27B
Input modalities
text, image
Output modalities
text
Modality
text+image->text
Native context
128K
Tokenizer
Gemini
Max output tokens
118K
Knowledge cutoff
2024-08
Source freshness
Metadata synced Oct 8, 2026, 10:17 PM UTC

Open-weight builds

Ollama variants

Every published Ollama tag matched to this model, including parameter editions, instruction/base variants, and quantized builds.

20 builds
VariantParametersTypeQuantizationSizeContextInput
gemma3:270m-it-bf16270M—BF16543 MB32Ktext
gemma3:270m-it-fp16270M—FP16543 MB32Ktext
gemma3:270m270M—Q8_0292 MB32Ktext
gemma3:270m-it-qat270M——241 MB32Ktext
gemma3:1b-it-fp161B—FP162 GB32Ktext
gemma3:1b1B—Q4_K_M806 MB32Ktext
gemma3:1b-it-q8_01B—Q8_01.1 GB32Ktext
gemma3:1b-it-qat1B——1 GB32Ktext
gemma3:4b-it-bf164B—BF168.6 GB128Ktext + image
gemma3:latest4B—Q4_K_M3.4 GB128Ktext + image
gemma3:4b-it-q8_04B—Q8_05 GB128Ktext + image
gemma3:4b-it-qat4B——4 GB128Ktext + image
gemma3:12b-it-bf1612B—BF1624 GB128Ktext + image
gemma3:12b12B—Q4_K_M8.2 GB128Ktext + image
gemma3:12b-it-q8_012B—Q8_013 GB128Ktext + image
gemma3:12b-it-qat12B——8.9 GB128Ktext + image
gemma3:27b-it-bf1627B—BF1655 GB128Ktext + image
gemma3:27b27B—Q4_K_M17 GB128Ktext + image
gemma3:27b-it-q8_027B—Q8_030 GB128Ktext + image
gemma3:27b-it-qat27B——18 GB128Ktext + image
Additional source evidence
Source model
google/gemma-3-27b-it
Source family
gemma
Last updated
2025-03-12
Knowledge cutoff
2024-08
Model context
131K
Model output limit
131K
Open weights
Yes
File attachments
Yes
Reasoning
No
Tool calling
Yes
Temperature control
Yes

Downloadable weights

Complete source record (JSON)
{
  "id": "google/gemma-3-27b-it",
  "name": "Gemma 3 27B IT",
  "description": "Largest open Gemma 3 instruction model for multilingual text generation and visual understanding",
  "family": "gemma",
  "last_updated": "2025-03-12",
  "knowledge": "2024-08",
  "attachment": true,
  "reasoning": false,
  "tool_call": true,
  "temperature": true,
  "modalities": {
    "input": [
      "text",
      "image"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 131072,
    "output": 131072
  },
  "release_date": "2025-03-12",
  "open_weights": true,
  "weights": [
    {
      "label": "Hugging Face",
      "url": "https://huggingface.co/google/gemma-3-27b-it"
    }
  ]
}

Infrastructure

Provider availability

All providers
ProviderEndpointContextInput /M tokensOutput /M tokensCached input /M tokensCache write /M tokensUptime 1dData handling
DeepInfradeepinfra/fp8131K$0.08$0.16——99.79%
ZDR: Yes

Training: Unknown

Prompt logging: Unknown

Retention: Unknown

Checked 2026-10-15

NextBitnextbit/fp8131K$0.08$0.35$0.04—99.65%
ZDR: Yes

Training: Unknown

Prompt logging: Unknown

Retention: Unknown

Checked 2026-10-15

Parasailparasail/fp8131K$0.08$0.45$0.04—99.09%
ZDR: Yes

Training: Unknown

Prompt logging: Unknown

Retention: Unknown

Checked 2026-10-15

Nebiusnebius/fp8110K$0.10$0.30——96.07%
ZDR: Yes

Training: Unknown

Prompt logging: Unknown

Retention: Unknown

Checked 2026-10-15

Novitanovita/bf1698K$0.12$0.20——96.22%
ZDR: Yes

Training: Unknown

Prompt logging: Unknown

Retention: Unknown

Checked 2026-10-15

Provider data:OpenRouter

Text provenance

Gemini app/web

Google SynthID text watermarking

Google DeepMind says text generated in the Gemini app and web experience is watermarked with SynthID by adjusting token probabilities.

Official explanation

Deployment graph

Can I run this?

Vertical Tabs only shows compatibility when it has structured evidence tying this exact model to hardware, runtime, quantization, memory use, and observed performance.

Browse hardware