Phi-2

By Microsoft, in the Phi family.

Category scores

Out of 100, from independent and third-party results only, by method category-2026-09-29.1.

Not enough independent results yet

Only independent and third-party results rank a model; self-reported and unattributed numbers never do.

Every result we hold for this model, with who ran it and where it was published.

Knowledge & reasoning results
Benchmark	Metric	Value	Provenance	Setting	Evaluated	Source
Adversarial NLI (ANLI)	Accuracy	42.5%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
AI2 Reasoning Challenge (ARC)	Challenge set accuracy	75.9%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
BIG-Bench Hard (BBH)	Average accuracy	59.4%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
HellaSwag	Accuracy	53.6%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
MMLU	Accuracy	58.4%	Third party	—	—	Stanford CRFM Leaderboard via Epoch AI
MMLU	Accuracy	56.3%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
OpenBookQA	Accuracy	73.6%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
OpenBookQA	Accuracy	73.6%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
WinoGrande	Accuracy	54.7%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI

Factuality & hallucination results
Benchmark	Metric	Value	Provenance	Setting	Evaluated	Source
TriviaQA	Exact match	45.2%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI