How much does the Mistral AI API cost?

It depends on the model. Across the 6 models tracked here, input ranges from $0.1500 to $1.50 per million tokens and output from $0.1500 to $7.50. On a typical chat turn of 1,500 input and 300 output tokens the range is $0.000270 to $0.004500 per request.

What is the cheapest Mistral AI model?

Ministral 3 8B, at $0.1500 per million input tokens and $0.1500 per million output. Whether it is cheap enough depends on whether it can do your task — the cost of finding out is usually a few cents.

How current is this Mistral AI pricing?

Every model entry stores the URL it was read from and the date it was read, both shown in the table above. The full data set was last reviewed on August 5, 2026. Prices change without notice, so verify against Mistral AI's own documentation before committing spend.

6 models tracked

Mistral AI API pricing

Mistral is the only provider here that publishes open weights for models it also sells through an API, which makes it the one place a build-versus-buy comparison can be made on the same model rather than on a substitute.

Models tracked: 6
Cheapest input: $0.1500
Highest input: $1.50
Exact token counts: 0 / 6

Every Mistral AI model, cheapest first

Mistral AI model pricing
Model	Input	Cached	Output	Context	Count	Chat turn
Ministral 3 8B	$0.1500	—	$0.1500	256K	Estimate	$0.000270
Mistral Small 4	$0.1500	—	$0.6000	256K	Estimate	$0.000405
Mistral Large 3	$0.5000	—	$1.50	256K	Estimate	$0.001200
Devstral 2	$0.4000	—	$2.00	256K	Estimate	$0.001200
Magistral Medium	$2.00	—	$5.00	256K	Estimate	$0.004500
Mistral Medium 3.5	$1.50	—	$7.50	256K	Estimate	$0.004500

Chat turn is 1,500 input and 300 output tokens. Prices read from Mistral AI’s pricing documentation; each row links to its own dated source.

What is distinctive about Mistral AI pricing

01
Several models are priced with input and output at the same rate, which is unusual. It makes output-heavy work — drafting, expansion, long answers — disproportionately cheap relative to providers that charge a premium on output.
02
A 90% cached input discount is advertised but not published per model, so no cached figure appears in the table above. An unlisted rate is not the same as a known one, and a calculator built on an inferred number is a calculator built on a guess.
03
Batch processing is discounted 50%, in line with the rest of the market.

Token counting on Mistral AI

Mistral publishes its Tekken tokenizer, but only as a Python package with no browser build, so counts here are estimates. If precision matters, tokenize from a backend with mistral-common rather than trusting a client-side figure.

Counts shown here for Mistral AI models are estimates and are labelled as such everywhere they appear. The methodology page documents every scaling factor used.

Compare with other providers

Or shortlist across all of them at once in the model finder, filtering by budget, context window and whether token counts can be exact.

Frequently asked questions

How much does the Mistral AI API cost?: It depends on the model. Across the 6 models tracked here, input ranges from $0.1500 to $1.50 per million tokens and output from $0.1500 to $7.50. On a typical chat turn of 1,500 input and 300 output tokens the range is $0.000270 to $0.004500 per request.
What is the cheapest Mistral AI model?: Ministral 3 8B, at $0.1500 per million input tokens and $0.1500 per million output. Whether it is cheap enough depends on whether it can do your task — the cost of finding out is usually a few cents.
Can I count Mistral AI tokens exactly?: Not in a browser. Mistral publishes its Tekken tokenizer, but only as a Python package with no browser build, so counts here are estimates. If precision matters, tokenize from a backend with mistral-common rather than trusting a client-side figure. Counts shown on this site for Mistral AI models are labelled as estimates everywhere they appear.
How current is this Mistral AI pricing?: Every model entry stores the URL it was read from and the date it was read, both shown in the table above. The full data set was last reviewed on August 5, 2026. Prices change without notice, so verify against Mistral AI's own documentation before committing spend.