Dolly V2 12B: benchmark results
Databricks' Dolly v2 12B, derived from EleutherAI's Pythia-12b and fine-tuned on about 15,000 employee-generated instruction-response records, released under a permissive license for commercial use. Provider: Databricks. Released 2023-04-12.
Aggregate Rating 1154 ± 12, rank #1620 of 1656 rated models, from 117 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLMsPark - Dictator Game (Multiple Play) | 1040.47 | Elo rating (initial 1000, K=32) | 92.3 |
| LLMsPark - Nim Game | 1180.35 | Elo rating (initial 1000, K=32) | 92.3 |
| LLMsPark - Trust Game (Multiple Play) | 1179.13 | Elo rating (initial 1000, K=32) | 92.3 |
| MMLU-by-task - Global Facts | 39 | Accuracy (%) | 85.7 |
| LLMsPark - Dictator Game (Single Play) | 1032.44 | Elo rating (initial 1000, K=32) | 76.9 |
| MMLU-by-task - Machine Learning | 34.82 | Accuracy (%) | 60.8 |
| MMLU-by-task - High School Mathematics | 26.67 | Accuracy (%) | 45.9 |
| ToolBench - WebShop Long | 0 | Task Score | 44.2 |
| MMLU-by-task - College Physics | 23.53 | Accuracy (%) | 41.6 |
| MMLU-by-task - Elementary Mathematics | 26.98 | Accuracy (%) | 39.2 |
| MMLU-by-task - HellaSwag | 54.63 | Accuracy (%) | 38.9 |
| LLMsPark - Prisoner's Dilemma (Multiple Play) | 906.54 | Elo rating (initial 1000, K=32) | 38.5 |
Interactive version: theaggregate.ai/model?slug=dolly-v2-12b · How It Works · Data refreshed daily, snapshot 2026-10-20.