The Model Index
Models/Microsoft

Phi-3

MicrosoftOpen weights

Phi-3 is an open-weights AI model from Microsoft with an index score of 97 (GPT-4 = 100), provisional.

9780–114Index scoreProvisional · results in fewer than three areas so far; not ranked yet
–Price per million tokensno public API price found
–Context windowtokens it reads at once
23 Apr 2024Released2 years 7 months ago

Where it sitson the frontier

801001201401601802002023202420252026GPT-4 at launch = 100Phi-3

GPT-4 reached this score 13 months earlier.

Test resultsby area · tick = best by any model

Reasoning & knowledgeevidence
GPQA Diamond (Epoch AI)
28%
Humanity's Last Exam
no result
Mathsevidence
FrontierMath (tiers 1-3)
no result
Codingno evidence yet
SWE-bench Verified
no result
Terminal-Bench 4.0
no result
Agentsno evidence yet
APEX-Agents
no result
Professional workno evidence yet
Finance Agent (v2)
no result
Novel problemsno evidence yet
ARC-AGI-2
no result
Long tasksno evidence yet
METR time horizon
no result

Third-party results only, at the model’s best standard setting. Hover a bar for who ran it.

What others sayrank on each leaderboard

Best attop three among current models

Not in the top three among current models on any category we track.

Facts and sources

Released23 Apr 2024
Announcementhuggingface.co ↗
Release entryPhi-3
Scored asphi-3-medium 14B
Output speed–
Hugging Facenot linked

Microsoft releasesaround this one

This page as a chartDownload PNG