benchr
Pricing Guide · Microsoft

Phi-4 API Pricing

Detailed token costs, context window limits, benchmark performance, and caching structures for Phi-4.

Last reviewed: June 6, 2026 Sourced from official docs

Phi-4 Cost Breakdown

Dimension Value / Cost
Input Tokens Cost / 1MSelf-hosted
Output Tokens Cost / 1MSelf-hosted
Cached Input Tokens Cost / 1M—
Max Context Window16,000 tokens
SWE-bench Verified Score30.0%
GPQA Diamond Score56.1%

Best suited for:

  • Local inference on consumer hardware
    • \n
  • Edge deployment
    • \n
  • Reasoning at tiny scale

← Back to all pricing guides