Microsoft·released 2024-04-231 source
phi-3-mini 3.8B benchmark scores: 9 benchmarks tracked, leading 1. 0% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 1 of 9 benchmarks·best result 84 on OpenBookQA (self-reported)·0 of 9 independently reproduced
Head-to-headphi-3-mini 3.8B vs phi-3-small 7.4BLeads on: OpenBookQA
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| OpenBookQA | 84accuracy (%) | self-reported· optimizedT1 | 2024-04-23Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| ARC AI2 | 79.87accuracy (%) | self-reported· optimizedT1 | 2024-04-23Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| HellaSwag | 68.93accuracy (%) | self-reported· optimizedT1 | 2024-04-23Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| TriviaQA | 64accuracy (%) | self-reported· optimizedT1 | 2024-04-23Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| BBH | 62.27accuracy (%) | self-reported· optimizedT1 | 2024-04-23Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| MMLU | 58.4accuracy (%) | self-reported· optimizedT1 | 2024-04-23Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| Winogrande | 41.6accuracy (%) | self-reported· optimizedT1 | 2024-04-23Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| ANLI | 29.2accuracy (%) | self-reported· optimizedT1 | 2024-04-23Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone |
| Chess Puzzles | 0accuracy (%) | unverifiedT2 | 2024-04-23 |
Claims drawn from cited facts, not live model generation.
Developed by Microsoft, this model is a compact 3.8-billion-parameter model released in 2024, a scale that favors efficiency and lower cost over the greater headroom and expense of larger systems.
3 cited facts
This model is scored on 9 tracked benchmarks and currently holds the top score on 1 of them. Its benchmark record is self-reported, meaning the numbers are vendor-claimed and not yet independently confirmed, so they should be read as claims rather than verified results.
3 cited facts
Its mean gap to the best recorded score across its tracked benchmarks is 25.41 points.
1 cited fact
2 facts cross-checked across data sources: 2 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: curated_param_counts, epoch_benchmarks
phi-3-mini 3.8B is an AI model developed by Microsoft, released 2024-04-23. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
phi-3-mini 3.8B has recorded scores on 9 benchmarks, each shown with its evidence status.
phi-3-mini 3.8B has recorded scores on 9 benchmarks — OpenBookQA, Winogrande, TriviaQA, MMLU, Chess Puzzles, ARC AI2, and 3 more. The full table above shows each score with its evidence status.
0 of 9 recorded scores (0%) are independently reproduced rather than self-reported by the lab.
phi-3-mini 3.8B has 9 tracked claims: 8 self-reported, 1 unverified.