Gemma 3 4B is 16.0× cheaper per million tokens (blended). Devstral 2 2512 has the larger context window (262,144 vs 131,072 tokens, 2.0×).
Gemma 3 4B vs Devstral 2 2512
Live catalog fields. Best-in-row is highlighted.
Gemma 3 4Bopen
Google · google__gemma-3-4b-it
Devstral 2 2512open
Mistral · mistralai__devstral-2512
Identity
| Field | google__gemma-3-4b-it | mistralai__devstral-2512 |
|---|---|---|
| Provider | Mistral | |
| Slug | google__gemma-3-4b-it | mistralai__devstral-2512 |
| Status | live | live |
| Open weights | Yes | Yes |
| License | — | — |
| HuggingFace | google/gemma-3-4b-it | mistralai/Devstral-2-123B-Instruct-2512 |
Quality
| Field | google__gemma-3-4b-it | mistralai__devstral-2512 |
|---|---|---|
| Quality index (0–100) | — | — |
Cost
| Field | google__gemma-3-4b-it | mistralai__devstral-2512 |
|---|---|---|
| Input $ / 1M tokens | $0.05 | $0.40 |
| Cached input $ / 1M | — | $0.04 |
| Output $ / 1M tokens | $0.10 | $2.00 |
Context
| Field | google__gemma-3-4b-it | mistralai__devstral-2512 |
|---|---|---|
| Max context tokens | 131,072 | 262,144 |
| Max output tokens | 16,384 | 209,715 |
Modalities
| Field | google__gemma-3-4b-it | mistralai__devstral-2512 |
|---|---|---|
| Input | text, image | text, file |
| Output | text | text |