Gemma 4 26B A4B vs Gemma 4 31B
| Spec | Gemma 4 26B A4B | Gemma 4 31B |
|---|---|---|
| Respan Index | 18.5 (#142) | 18.7 (#140) |
| Blended price per 1M | - | - |
| Context window | 256K | 256K |
| Input / output per 1M | $0 / $0 | $0 / $0 |
| Max output | - | - |
| Input | text, vision | text, vision |
| Released | 2026-04-02 | 2026-04-02 |
Respan Index by category
Each category is scored 0 to 100, and how far ahead the winner is. Leads under a point are close to even.
| Category | Gemma 4 26B A4B | Gemma 4 31B |
|---|---|---|
| Coding | 4.1 Gemma 4 26B A4B | 7.8 Gemma 4 31B |
| Agents | 13.4* Gemma 4 26B A4B | 12.5* Gemma 4 31B |
| Reasoning | 20.4 Gemma 4 26B A4B | 19.7 Gemma 4 31B |
| Knowledge & language | 15.3 Gemma 4 26B A4B | 11.0 Gemma 4 31B |
| Human preference | 50.0 Gemma 4 26B A4B | 50.4 Gemma 4 31B |
* Estimated: no published scores in that category. How the Respan Index works
Benchmarks both report
Gemma 4 31B leads on 9 of the 11 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | Gemma 4 26B A4B | Gemma 4 31B |
|---|---|---|
| LMArena Elo | 1438 | 1452 |
| LMArena Vision (LMArena) | 1241 | 1265 |
| LMArena WebDev (LMArena) | 1358 | 1370 |
| Chess Puzzles (Epoch AI) | 6% | 5% |
| GPQA Diamond | 82.3% | 84.3% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 82.2% | 73.3% |
| WeirdML (Håvard Tveit Ihle) | 35.2% | 52.3% |
| AIME 2026 | 88.3% | 89.2% |
| Humanity's Last Exam (no tools) | 8.7% | 19.5% |
| LiveCodeBench v6 | 77.1% | 80% |
| MMLU-Pro | 82.6% | 85.2% |
FAQ
Is Gemma 4 26B A4B better than Gemma 4 31B?
Gemma 4 31B scores 18.7 on the Respan Index (#140), ahead of Gemma 4 26B A4B at 18.5 (#142). Gemma 4 31B leads on 9 of the 11 benchmarks both report.
Which has the larger context window?
Both accept 256K tokens of context.