Gemma 4 26B A4B: benchmark results
Google's Apache-2.0 Gemma 4 MoE (April 2026) with 26B total and ~4B active parameters, multimodal with context up to 256K. Provider: Google. Released 2026-04-02. Access: Open.
Unified ELO 1571 ± 1, rank #357 of 1392 rated models, from 354 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Kaggle FACTS Grounding | 80.87 | Score (%) | 100 |
| EuroEval Bulgarian NLU - Cinexio | 57.15 | Sentiment classification Score (%) | 92.4 |
| EuroEval Icelandic NLU - NQII | 57.2 | Reading comprehension Score (%) | 90.6 |
| MathVision | 82.4 | Overall Accuracy (%) | 88.8 |
| Chatbot Arena (Text - Math) | 1468 | Arena Score | 88.5 |
| EuroEval Estonian NLU - MultiWikiQA ET | 70.19 | Reading comprehension Score (%) | 87.1 |
| EuroEval Czech NLU - SQAD | 68.08 | Reading comprehension Score (%) | 85.5 |
| EuroEval Ukrainian NLU - MultiWikiQA UK | 62.69 | Reading comprehension Score (%) | 85.4 |
| EuroEval German NLU - Germanquad | 62.82 | Reading comprehension Score (%) | 83.8 |
| Chatbot Arena (Text - Instruction Following) | 1439 | Arena Score | 83.1 |
| EuroEval Polish NLU - Polemo2 | 91.51 | Sentiment classification Score (%) | 82.8 |
| EuroEval Serbian NLU - MultiWikiQA SR | 62.56 | Reading comprehension Score (%) | 81.9 |
Interactive version: theaggregate.ai/model?slug=gemma-4-26b-a4b · How It Works · Data refreshed daily, snapshot 2026-09-05.