Phi-4
Phi-4 is an open-weights AI model from Microsoft with an index score of 114 (GPT-4 = 100), rank 166 of 217.
11498–130Index score#166 of 217 models
$0.070/ $0.14Price per million tokensinput / output · via OpenRouter
16KContext windowtokens it reads at once
12 Dec 2024Released1 year 11 months ago
Where it sitson the frontier
o1-mini reached this score 3 months earlier.
Test resultsby area · tick = best by any model
Reasoning & knowledgeevidence
Mathsevidence
Codingno evidence yet
Agentsevidence
Professional workno evidence yet
Novel problemsno evidence yet
Long tasksno evidence yet
Third-party results only, at the model’s best standard setting. Hover a bar for who ran it.
What others sayrank on 8 leaderboards
The Model Indexthird-party tests, our method#166of 217 · 114
Epoch Capabilities Indexoverall capability from many benchmarks#143of 231 · 129.2LMArena · Mathsmaths questions in chat#235of 346 · 1265LMArena · Codingcoding questions in chat#256of 357 · 1306LMArena · Instruction followingfollowing detailed instructions#264of 362 · 1245LMArena · Creative writingcreative writing#267of 360 · 1210LMArena · Textoverall chat quality, judged by people#269of 362 · 1256LMArena · Long promptslong, complex requests#263of 340 · 1267Best attop three among current models
Not in the top three among current models on any category we track.
Where to use itprices checked 18 Nov 2026
API providers 1 · per 1M tokens, in / out
Run it yourself
Official weightsmicrosoft/phi-4↗Host prices from OpenRouter; going direct can differ.
Facts and sources
Released12 Dec 2024
Announcementnot linked yet
Release entryPhi-4
Scored asPhi-4
Output speed–
Hugging Facemicrosoft/phi-4 ↗