h2o-danube-1.8B-chat: benchmark results
H2O.ai's h2o-danube-1.8b-chat, a 1.8B chat model using an adjusted Llama 2 architecture with Mistral-style sliding window attention and a 16,384-token context. Provider: H2O.ai. Released 2024-01-25. Access: Open.
Aggregate Rating 1257 ± 30, rank #1506 of 1635 rated models, from 17 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MuSR | 39.89 | Score | 41 |
| Open LLM Leaderboard v1 - GSM8K | 17.36 | Accuracy (%) (5-shot) | 38.9 |
| Open Portuguese LLM - FaQuAD NLI | 44.71 | Macro F1 (%) | 26.5 |
| Open Portuguese LLM - ENEM | 23.09 | Accuracy (%) | 19.3 |
| Open LLM Leaderboard v1 - HellaSwag | 68.06 | Normalized accuracy (%) (10-shot) | 19 |
| Open LLM Leaderboard v1 - WinoGrande | 65.35 | Accuracy (%) (5-shot) | 18.6 |
| Open LLM Leaderboard v1 - MMLU | 33.41 | Accuracy (%) (5-shot) | 17.7 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 41.64 | MC2 (%) (0-shot) | 17.7 |
| Open LLM Leaderboard v1 - ARC Challenge | 41.13 | Normalized accuracy (%) (25-shot) | 17.4 |
| Open Portuguese LLM - ASSIN2 RTE | 51.34 | Macro F1 (%) | 16.5 |
| Open LLM Leaderboard - IFEval | 21.99 | Score | 14.9 |
| Open LLM Leaderboard - BBH | 32.2 | Score | 12.5 |
Interactive version: theaggregate.ai/model?slug=h2o-danube-1-8b-chat · How It Works · Data refreshed daily, snapshot 2026-11-05.