ϕ

Phi-4

by Microsoft

4.2
Benchmark RatingLaunched 2024Free / open weights

Phi-4 is Microsoft's compact 14B-parameter model that punches well above its weight on reasoning and math. Trained heavily on synthetic data, it's ideal for edge deployment and cost-sensitive apps.

14B paramsSynthetic-trainedReasoning
Visit Site

Pricing Tiers

Click a tier to see full details

Open weights — $0

Subscribe

Self-host, free

Plan: Open weightsPrice: $0Tool: Phi-4

Capabilities & Specs

Best for
Efficient reasoning
Context window
16K
Image generation
Voice mode
Code execution
Web browsing
Offline / open weights
Public API
Starting price
Free / open

Technical Benchmarks

Raw scores from public leaderboards (LMArena, OpenAI, Anthropic). Higher is better.

MMLU79%

Language Understanding

HumanEval76%

Code Generation

GSM8K88%

Math Reasoning

GPQA35%

Graduate QA

LMArena ELO1185

Chatbot Arena

Pros

  • Excellent reasoning per param
  • Runs on modest hardware
  • Free + open

Cons

  • Small context
  • Limited general knowledge

Tasks Phi-4 Covers

WritingCodingData