Smaug-34B-v0.1: benchmark results
Abacus.AI's DPO-Positive fine-tune of Jon Durbin's Bagel 34B, a Yi-34B-lineage model. Provider: Abacus.AI. Released 2024-01-25. Access: Open.
Aggregate Rating 1519 ± 23, rank #723 of 1656 rated models, from 25 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - ARC Challenge | 74.23 | Normalized accuracy (%) (25-shot) | 99.1 |
| Open LLM Leaderboard v1 - MMLU | 76.66 | Accuracy (%) (5-shot) | 97.8 |
| Open LLM Leaderboard v1 - GSM8K | 72.18 | Accuracy (%) (5-shot) | 97.3 |
| Open Chinese LLM - ARC Challenge | 65.87 | Accuracy (%) | 96.9 |
| Open Chinese LLM - WinoGrande | 71.59 | Accuracy (%) | 96.9 |
| Open Chinese LLM - C-Eval Semantic | 90.75 | Accuracy (%) | 96.6 |
| Open Chinese LLM Leaderboard | 70.61 | Official Nonzero-Task Mean (%) | 96.2 |
| Open Chinese LLM - HellaSwag | 71.11 | Accuracy (%) | 95.9 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 70.22 | MC2 (%) (0-shot) | 90.2 |
| Open LLM Leaderboard v1 - WinoGrande | 83.66 | Accuracy (%) (5-shot) | 90.1 |
| Open Portuguese LLM - FaQuAD NLI | 80.52 | Macro F1 (%) | 90.1 |
| Open Portuguese LLM - BLUEX | 65.79 | Accuracy (%) | 87.3 |
Interactive version: theaggregate.ai/model?slug=smaug-34b-v0-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.