Gemma 3 27B (IT): benchmark results
Instruction-tuned Gemma 3 27B checkpoint. Provider: Google. Released 2025-03-12. Access: Open.
Aggregate Rating 1576 ± 3, rank #523 of 1656 rated models, from 2913 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Bengali Dialect QA Consistency | 8.71 | Macro-average over the nine dialects of the Dialect consiste | 100 |
| Bengali Dialect QA Consistency - Kishoreganj | 9.3 | Dialect consistency score (0-10): a Gemini 2.5 Flash judge ( | 100 |
| Bengali Dialect QA Consistency - Narail | 8.38 | Dialect consistency score (0-10): a Gemini 2.5 Flash judge ( | 100 |
| Bengali Dialect QA Consistency - Noakhali | 9.03 | Dialect consistency score (0-10): a Gemini 2.5 Flash judge ( | 100 |
| Bengali Dialect QA Consistency - Sylhet | 8.85 | Dialect consistency score (0-10): a Gemini 2.5 Flash judge ( | 100 |
| COMPASS Policy - Denied Base | 39.64 | Policy Alignment Score (%) | 100 |
| DementiaCare-Bench - 128 Frames | 85.1 | Accuracy (%) | 100 |
| DementiaCare-Bench - 16 Frames | 84.1 | Accuracy (%) | 100 |
| DementiaCare-Bench - 32 Frames | 84.7 | Accuracy (%) | 100 |
| DementiaCare-Bench - 48 Frames | 84.9 | Accuracy (%) | 100 |
| DementiaCare-Bench - 64 Frames | 85.1 | Accuracy (%) | 100 |
| DementiaCare-Bench - 8 Frames | 83.5 | Accuracy (%) | 100 |
Interactive version: theaggregate.ai/model?slug=gemma-3-27b-it · How It Works · Data refreshed daily, snapshot 2026-10-20.