UltraLM-13B: benchmark results
A chat fine-tune of the original LLaMA-13B on UltraChat, distributed as delta weights that must be recovered against the base model. Delta tensors are not a stand-alone checkpoint. Provider: OpenBMB. Released 2023-06-27. Access: Open.
Aggregate Rating 1138 ± 45, rank #10163 of 10916 rated models, from 8 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AlpacaEval 1.0 | 80.64 | Win Rate (%) | 42.6 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 48.61 | MC2 (%) (0-shot) | 41.9 |
| AlpacaEval 2.0 | 7.11 | LC Win Rate (%) | 11.7 |
| Open LLM Leaderboard v1 - ARC Challenge | 29.44 | Normalized accuracy (%) (25-shot) | 8.6 |
| Open LLM Leaderboard v1 - GSM8K | 0 | Accuracy (%) (5-shot) | 4.5 |
| Open LLM Leaderboard v1 - HellaSwag | 25.99 | Normalized accuracy (%) (10-shot) | 2.2 |
| Open LLM Leaderboard v1 - WinoGrande | 49.57 | Accuracy (%) (5-shot) | 1.7 |
| Open LLM Leaderboard v1 - MMLU | 23.12 | Accuracy (%) (5-shot) | 1 |
Interactive version: theaggregate.ai/model?slug=ultralm-13b · How It Works · Data refreshed daily, snapshot 2026-11-08.