Language model rankings / Gemma 4 31B IT
Gemma 4 31B IT
Google DeepMind
- Marsel score
- 82.2 (2/4 inputs)
- Licence
- Open weightsApache 2.0
- Context
- 262K tokens
- Modalities
- —
Released 2026-04-02
Measurements
Each figure is a measurement published by the named source on the named date. Where the source publishes its own page for this model, the date links to it so the figure can be checked at origin.
| Metric | Value | Source | Measured | In score |
|---|---|---|---|---|
| Human preference (Elo)A rating derived from blind head-to-head votes by people. It measures which answer humans preferred — not correctness. | 1452 | LMArena | 2026-10-15 ↗ | yes |
| ReasoningPerformance on the source’s reasoning evaluation set, on its own scale. | 75.8 | Epoch AI | 2026-08-06 ↗ | yes |
Price: $0.20 in / $0.40 out per 1M tokens · OpenRouter · effective 2026-11-10
Price history
as of 2026-11-10Output price per 1M tokens. Stepped — a price holds until it is changed.
Stepped: a price holds until it is changed, so the line does not slope between announcements.
Showing 1 model.
View data
| Effective | Output $/1M |
|---|---|
| 2026-08-19 | $0.34 |
| 2026-08-20 | $0.34 |
| 2026-08-21 | $0.34 |
| 2026-08-22 | $0.34 |
| 2026-08-23 |
Human preference (Elo) over time
as of 2026-10-15Points are real observations. Gaps are gaps — nothing is interpolated.
Showing 1 model.
View data
| Measured | Human preference (Elo) |
|---|---|
| 2026-08-12 | 1451 |
| 2026-08-19 | 1451 |
| 2026-08-21 | 1451 |
| 2026-08-27 | 1451 |
| 2026-09-01 | 1451 |
| 2026-09-13 | 1451 |