Back to Models
microsoft logo
microsoft/phi-4
Not Available

Phi 4

[Microsoft Research](/microsoft) Phi-4 is designed to perform well in complex reasoning tasks and can operate efficiently in situations with limited memory or where quick responses are needed. At 14 billion parameters, it was trained on a mix of high-quality synthetic datasets, data from curated websites, and academic materials. It has undergone careful improvement to follow instructions accurately and maintain strong safety standards. It works best with English language inputs. For more information, please see [Phi-4 Technical Report](https://arxiv.org/pdf/2412.08905)

1/10/2025
16,384 tokens
#214 Text (Japanese)

Specifications

Modalities

Input
text
Output
text

Supported Parameters

frequency_penalty
max_tokens
min_p
presence_penalty
repetition_penalty
response_format
seed
stop
structured_outputs
temperature
top_k
top_p

Max Output Tokens

16,384

Frequently asked questions

Is Phi 4 available on FastMetal?
Not at the moment. GLM 5.3 Flash, from the same lab, is available on the FastMetal API today.
What is the context window of Phi 4?
16,384 tokens, shared between the prompt and the response.
How does Phi 4 rank?
#214 on the public arena's Japanese board (ELO 1,194). Ranks move as the leaderboard is updated.

Leaderboard

Text
OverallELO: 1,256
#304
JapaneseELO: 1,194
#214
ChineseELO: 1,267
#288
KoreanELO: 1,196
#227
EnglishELO: 1,272
#301
frenchELO: 1,266
#235
germanELO: 1,256
#234
spanishELO: 1,269
#234
russianELO: 1,246
#296
CodingELO: 1,306
#290
MathELO: 1,265
#268
Creative WritingELO: 1,210
#304
Instruction FollowingELO: 1,245
#298
Hard PromptsELO: 1,277
#293
Multi-TurnELO: 1,242
#298