MicrosoftPhi 4Open SourceArena #209Jan 10, 2025
Microsoft: Phi 4
Microsoft Research 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
Context Window
16K
tokens
Max Output
16K
tokens
Released
Jan 10, 2025
Arena Rank
#209
of 305 models
Capabilities
👁Vision
🧠Reasoning
🔧Tool Calling
⚡Prompt Caching
🖥Computer Use
🎨Image Generation
Supported Parameters
Frequency Penalty
Reduce repetition
Max Tokens
Output length limit
min_p
Presence Penalty
Encourage new topics
Repetition Penalty
Penalize repeated tokens
Response Format
JSON mode / structured output
Seed
Deterministic outputs
Stop Sequences
Custom stop tokens
structured_outputs
Temperature
Controls randomness
Top K
Top-K sampling
Top P
Nucleus sampling
Pricing Comparison
| Router | Input / 1M | Output / 1M | Cached Input / 1M |
|---|---|---|---|
| OpenRouter | $0.06 | $0.14 | — |
| Martian | $0.06 | $0.14 | — |
| DeepInfra | $0.07 | $0.14 | — |
Benchmarks
Open LLM Leaderboard
AverageOpen LLM Leaderboard
40.73/100IFEvalOpen LLM Leaderboard
68.82/100BBHOpen LLM Leaderboard
55.25/100MATH Lvl 5Open LLM Leaderboard
50/100GPQAOpen LLM Leaderboard
11.52/100MUSROpen LLM Leaderboard
10.13/100MMLU-PROOpen LLM Leaderboard
48.65/100Model IDs
OpenRouter
microsoft/phi-4Hugging Facemicrosoft/phi-4 ↗