Skip to content
Sovyron

Mistral Medium 3

mistral-medium131k context131k max outputreleased May 7, 2025Tool calling

As of Oct 4, 2026, the cheapest Mistral Medium 3 API price is $0.4 per 1M input tokens (Azure) and $2 per 1M output tokens (Azure). Mistral Medium 3 is offered by 9 providers in the mistral-medium family. Mistral's own rate is $2 per 1M output tokens. One provider lists it as a free-tier offer.

Cheapest input
$0.4
Cheapest output
$2
Official (Mistral)
$2
output per 1M tokens
Providers
9
offering this model
Spread
1.0×
max / min output
Free offers
1
free-tier offer

API pricing by provider

Metered per-token prices. Cheapest input and output are highlighted; † marks a context-tier surcharge. How prices are ranked.

Mistral Medium 3 API prices by provider in USD per 1M tokens, updated Oct 4, 2026
AzureCheapest$0.4Cheapest$2——128k
Azure Cognitive Services$0.4$2——128k
Eden AI$0.4$2——131k
Merge Gateway$0.4$2$0.04—128k
Mistral$0.4$2——131k
NanoGPT$0.4$2$0.2—131k
OpenRouter$0.4$2$0.04—131k
Pioneer$0.4$2$0.4$0.4128k
Nvidiafreefreefree——131k

Mistral Medium 3 pricing FAQ

What is the cheapest Mistral Medium 3 API?

As of Oct 4, 2026, Azure has the lowest Mistral Medium 3 output price at $2 per 1M tokens, and Azure has the lowest input price at $0.4 per 1M tokens. Prices are metered per-token rates in USD; free tiers and subscription plans are excluded.

How much does Mistral Medium 3 cost on Mistral?

Mistral charges $0.4 per 1M input tokens and $2 per 1M output tokens for Mistral Medium 3.

How many providers offer Mistral Medium 3?

9 providers list Mistral Medium 3 on Sovyron; 8 of them sell it at a metered per-token price.

Is Mistral Medium 3 free?

1 provider(s) list a free-tier offer: Nvidia. Free tiers usually have rate limits.

What is the context window of Mistral Medium 3?

Mistral Medium 3 supports a 131k-token context window and up to 131k output tokens.

Token cost calculator

What a month costs at one provider's published rates. Enter the tokens you expect to send and receive; cached tokens use the cached-input price where the provider lists one, and count as ordinary input where it does not.

Estimated

—