Model comparison
Gemma 3 vs Gemma 2
- Input
- $0.08
- Output
- $0.16
- Context
- 128K
- Released
- 2025-03-12
- Input
- Free to download and self-host
- Output
- Free to download and self-host
- Context
- 8,192 tokens
- Released
- 2024-06-27
Gemma 3 scores higher on 3 of the 3 benchmarks both models report; Gemma 3 has the larger context window (128K vs 8K).
Specs and pricing side by side
| Spec | Gemma 3 | Gemma 2 |
|---|---|---|
| Maker | Googlesame | |
| Released | 2025-03-12 | 2024-06-27 |
| Context window | 128K | 8,192 tokens |
| Max output | 8K | Not separately specified; output shares the… |
| Input price | $0.08 | Free to download and self-host |
| Output price | $0.16 | Free to download and self-host |
| Parameters | 1B · 4B · 12B · 27B (270M added Aug 2025) | Three sizes: 2B (2.6B), 9B, and 27B parameters |
| License | Gemma Terms of Use | Gemma (Gemma Terms of Use) — a custom, commercially permissive license that allows redistribution, fine-tuning, commercial use, and derivative works subject to a prohibited-use policy. Not OSI-approved (not Apache 2.0). |
| Weights | Open weights | Open weightssame |
| Modalities | Text, Vision | text |
| Knowledge cutoff | August 2024 | Not officially published by Google for Gemma 2 |
Benchmarks both models report
| Benchmark | Gemma 3 | Gemma 2 |
|---|---|---|
| GSM8K | 95.9% | 74% |
| MATH | 89% | 42.3% |
| HumanEval | 87.8% | 51.8% |
Scores are as published by each maker or its cited source; harness, reasoning effort and date can differ between rows, so treat small gaps as ties. Sources are listed on each model page.
Which one should you pick?
Pick Gemma 3 if…
- You care about GSM8K, MATH, HumanEval (it scores higher on 3 shared benchmarks).
- You need the longer context (128K vs 8K).
- You need Text / Vision input, which Gemma 2 lacks.
- Your use case is: self-hosted chat and assistants where data must stay local.
Pick Gemma 2 if…
- You need text input, which Gemma 3 lacks.
- Your use case is: on-device and self-hosted text generation, summarization, and extraction where data must stay local.
FAQ
Which is better, Gemma 3 or Gemma 2?
Gemma 3 scores higher on 3 of the 3 benchmarks both models report; Gemma 3 has the larger context window (128K vs 8K). The better pick depends on whether you value benchmark scores, price per token, context length or openness for your workload.
What are the context windows of Gemma 3 and Gemma 2?
Gemma 3 offers 128K with up to 8K output; Gemma 2 offers 8,192 tokens with up to Not separately specified; output shares the 8,192-token context window output.
Is Gemma 3 newer than Gemma 2?
Gemma 3 was released 2025-03-12; Gemma 2 was released 2024-06-27.