Microsoft
Efficient

Phi-3.5 Mini

Compact

The 3.8B Phi-3.5. Designed for on-device reasoning with strong instruction following.

Phi-3.5 Mini is a efficient AI model from Microsoft. It costs $0.1300 per million input tokens and $0.5200 per million output tokens (blended $0.2470/M), with a 128,000-token context window.

RatesUSD per million tokens. Blend is 70% input, 30% output, our calculation.

INPUT
$0.1300/M
per million input tokens
OUTPUT
$0.5200/M
per million output tokens
BLENDED 70/30
$0.2470/M
unchanged since 3 May
CONTEXT
128,000
tokens

Hand-maintained from Microsoft's published pricing; no check date on record.

What it is good at
  • On-device capable
  • Open weights (MIT)
  • 128K context
Typical use cases
  • Edge assistants
  • On-device summarisation

Benchmarks

vs. best public score
MMLU69%
Multitask academic knowledge across 57 subjects. Source: Microsoft (self-reported), MMLU (5-shot, Phi-3.5-mini-instruct), read 2026-11-10.
MATH48.5%
High-school competition math problems. Source: Microsoft (self-reported), MATH (0-shot, CoT), read 2026-11-10.
Python function synthesis from docstrings. No source recorded yet.
Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources, treat as directional rather than authoritative.

More from Microsoft

See all 8 →

Frequently asked questions

How much does Phi-3.5 Mini cost?

Phi-3.5 Mini costs $0.1300 per million input tokens and $0.5200 per million output tokens, for a blended reference rate of $0.2470 per million tokens.

What is Phi-3.5 Mini's context window?

Phi-3.5 Mini supports up to 128,000 tokens of context in a single request.

What is Phi-3.5 Mini best for?

Phi-3.5 Mini is well suited to On-device capable, Open weights (MIT) and 128K context.

Who makes Phi-3.5 Mini?

Phi-3.5 Mini is developed and served by Microsoft. It was released in Aug 2024.

Terms used on this page