This model is 22 months old. The same company has since released something newer: Decision-1 (2026-10-10). The details below are still accurate for this model — it just is not what we would recommend today.
Phi-4
Microsoft🇺🇸 United States
Microsoft's 14B parameter model that punches above its weight on STEM benchmarks.
Context window16K tokens
Input / 1M tokens$0.13
Output / 1M tokens$0.5
Version History
phi-4major
Phi-4 debuts as a 14B model that punches far above its weight on GPQA, MATH, and HumanEval. MIT license.