Gemma 3 Models by Google DeepMind
Last refreshed 2026-05-22. Next refresh: weekly.
Details
Capabilities
About
Gemma 3 is Google DeepMind’s open-weight Gemma generation, built from the same research line as Gemini 2.0 and sized at 1B, 4B, 12B, and 27B so it can run on a single GPU or TPU and on smaller devices. Larger sizes add image and short-video understanding alongside text; the family supports a long context window and wide multilingual coverage. Official quantized builds are published for lighter local and edge deployment.
Decision facts
- Best fit
- safetyvision and multimodal workJSON / Tool use
- Capability starting point
- ShieldGemma 2 with JSON / Tool use, structured outputs, and multimodal inputs
- Lowest tracked input
- Together AI - Gemma 3n-e4B · $0.020/1M · Together AI
- Closest related family
- DiffusionGemma
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
Use when the workload needs 8K context, 4B parameters, and JSON / Tool use.
Use when the workload needs 33K context, 12B parameters, and structured outputs.
Use when the workload needs 128K context, 27B parameters, and structured outputs.
Use when the workload needs 128K context, 4B parameters, and structured outputs.
Use when the workload needs 8K context and structured outputs.
Use when the workload needs 131K context, 27B parameters, and structured outputs.
Use when the workload needs 32K context and structured outputs.
Use when the workload needs 128K context and 12B parameters.
Use when the workload needs 32K context and 1B parameters.
Use when the workload needs 128K context and 27B parameters.
Use when the workload needs 128K context and 4B parameters.
Use when the workload needs safety, 4B parameters, and JSON / Tool use.
Use when the workload needs 4B parameters, JSON / Tool use, and structured outputs.
Use when the workload needs 400M parameters, JSON / Tool use, and structured outputs.
Use when the workload needs 2B parameters, JSON / Tool use, and structured outputs.
Use when the workload needs 3B parameters, JSON / Tool use, and structured outputs.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Together AI - Gemma 3n-e4B | Use when the workload needs 8K context, 4B parameters, and JSON / Tool use. | 2026-03 | 8K context4B parametersJSON / Tool use | Current |
| Gemma 3 12B | Use when the workload needs 33K context, 12B parameters, and structured outputs. | 2026-01 | 33K context12B parametersstructured outputs | Current |
| Gemma 3 27B PT | Use when the workload needs 128K context, 27B parameters, and structured outputs. | 2026-01 | 128K context27B parametersstructured outputs | Current |
| Gemma 3 4B IT | Use when the workload needs 128K context, 4B parameters, and structured outputs. | 2026-01 | 128K context4B parametersstructured outputs | Current |
| Gemma 3n 2B (free) | Use when the workload needs 8K context. | 2025-04 | 8K context | Current |
| Gemma 3n 4B (free) | Use when the workload needs 8K context and structured outputs. | 2025-04 | 8K contextstructured outputs | Current |
| Gemma 3 | Use when the workload needs structured outputs. | 2025-03 | structured outputs | Current |
| Gemma 3 27B | Use when the workload needs 131K context, 27B parameters, and structured outputs. | 2025-03 | 131K context27B parametersstructured outputs | Current |
| Gemma 3n | Use when the workload needs 32K context and structured outputs. | 2025-03 | 32K contextstructured outputs | Current |
| Gemma 3 12B Instruct | Use when the workload needs 128K context and 12B parameters. | 2025-01 | 128K context12B parameters | Current |
| Gemma 3 1B Instruct | Use when the workload needs 32K context and 1B parameters. | 2025-01 | 32K context1B parameters | Current |
| Gemma 3 27B Instruct | Use when the workload needs 128K context and 27B parameters. | 2025-01 | 128K context27B parameters | Current |
| Gemma 3 4B Instruct | Use when the workload needs 128K context and 4B parameters. | 2025-01 | 128K context4B parameters | Current |
| ShieldGemma 2 | Use when the workload needs safety, 4B parameters, and JSON / Tool use. | 2024-09 | safety4B parametersJSON / Tool use | Current |
| MedGemma | Use when the workload needs 4B parameters, JSON / Tool use, and structured outputs. | 2024-07 | 4B parametersJSON / Tool usestructured outputs | Current |
| MedSigLIP | Use when the workload needs 400M parameters, JSON / Tool use, and structured outputs. | 2024-07 | 400M parametersJSON / Tool usestructured outputs | Current |
| TxGemma | Use when the workload needs 2B parameters, JSON / Tool use, and structured outputs. | 2024-06 | 2B parametersJSON / Tool usestructured outputs | Current |
| PaliGemma | Use when the workload needs 3B parameters, JSON / Tool use, and structured outputs. | 2024-03 | 3B parametersJSON / Tool usestructured outputs | Current |
Release Timeline
9 release groupsSpecifications(18 models)
| Model | Released | Context | Parameters | Vision | Multimodal | JSON / Tool use | Structured Outputs |
|---|---|---|---|---|---|---|---|
| Together AI - Gemma 3n-e4B | 2026-03 | 8K | 4B | No | No | Yes | Yes |
| Gemma 3 12B | 2026-01 | 33K | 12B | No | No | No | Yes |
| Gemma 3 27B PT | 2026-01 | 128K | 27B | No | No | No | Yes |
| Gemma 3 4B IT | 2026-01 | 128K | 4B | No | No | No | Yes |
| Gemma 3n 2B (free) | 2025-04 | 8K | 5B (2B effective active) | No | No | No | No |
| Gemma 3n 4B (free) | 2025-04 | 8K | 8B (4B effective active) | No | No | No | Yes |
| Gemma 3 | 2025-03 | — | — | No | No | No | Yes |
| Gemma 3 27B | 2025-03 | 131K | 27B | No | No | No | Yes |
| Gemma 3n | 2025-03 | 32K | — | No | No | No | Yes |
| Gemma 3 12B Instruct | 2025-01 | 128K | 12B | No | No | No | No |
| Gemma 3 1B Instruct | 2025-01 | 32K | 1B | No | No | No | No |
| Gemma 3 27B Instruct | 2025-01 | 128K | 27B | No | No | No | No |
| Gemma 3 4B Instruct | 2025-01 | 128K | 4B | No | No | No | No |
| ShieldGemma 2 | 2024-09 | — | 4B | Yes | Yes | Yes | Yes |
| MedGemma | 2024-07 | — | 4B | Yes | Yes | Yes | Yes |
| MedSigLIP | 2024-07 | — | 400M | Yes | Yes | Yes | Yes |
| TxGemma | 2024-06 | — | 2B | No | No | Yes | Yes |
| PaliGemma | 2024-03 | — | 3B | Yes | Yes | Yes | Yes |
Available From(9 providers)
Pricing
Popular comparisons in this family
- Together AI - Gemma 3n-e4B vs Hunyuan Hy3 Preview224
- Gemma 3 vs Nemotron 3 Nano Omni221
- Together AI - Gemma 3n-e4B vs Phi-4 Mini Flash Reasoning98
- Gemma 3 12B vs Qwen3.5-9B86
- Gemma 3 12B vs Llama 3.1 Swallow 8B Instruct84
- Together AI - Gemma 3n-e4B vs Granite Guardian 3.0 8B62
- Gemma 3 12B vs Nemotron 3 Content Safety48
- Gemma 3 12B vs Phi-4 Mini Flash Reasoning44
- Together AI - Gemma 3n-e4B vs Nemotron Mini 4B Instruct41
- Together AI - Gemma 3n-e4B vs Xiaomi MiMo-V2.5-Pro38
Models(18)
Together AI - Gemma 3n-e4B
Gemma 3 12B
Gemma 3 27B PT
Gemma 3 4B IT
Gemma 3n 2B (free)
Gemma 3n 4B (free)
Gemma 3
Gemma 3 27B
Gemma 3n
Gemma 3 12B Instruct
Gemma 3 1B Instruct
Gemma 3 27B Instruct
Gemma 3 4B Instruct
ShieldGemma 2
MedGemma
MedSigLIP
TxGemma
PaliGemma