Microsoft: Phi 4 vs Mistral Small (latest)

Microsoft: Phi 4 is cheaper than Mistral Small (latest) by about 67% on a blended (75% input / 25% output) basis. Microsoft: Phi 4 charges $0.070000 / $0.140000 per 1M input/output tokens; Mistral Small (latest) charges $0.150000 / $0.600000.

FieldMicrosoft: Phi 4Mistral Small (latest)
Input $/Mtok$0.070000$0.150000
Output $/Mtok$0.140000$0.600000
Cache read $/Mtok–$0.015000
Cache write $/Mtok––
Input $/Mcharn/a0.053343
Output $/Mcharn/a0.213371
Context window16384256000
Tokenizer–mistral

Price history

Same workload, monthly cost

Illustrative monthly cost at 10M tokens/month, split per workload preset. Not a substitute for the calculator, which uses your own volume, cache and batch assumptions.

WorkloadMicrosoft: Phi 4Mistral Small (latest)
Coding agent$0.74$1.73
RAG chat$0.81$2.17
Summarisation (batch)$0.77$1.95
Classification$0.71$1.59
Chat assistant$0.98$3.30
Content generation$1.26$5.10

Cite this page

Microsoft: Phi 4 vs Mistral Small (latest): $0.070000/$0.140000 vs $0.150000/$0.600000 per 1M tokens, as of 2026-10-04 (Token Barometer, https://tokenbarometer.com/compare/microsoft_phi-4-vs-mistral_mistral-small-2503).

Microsoft: Phi 4 page | Mistral Small (latest) page | Open in calculator