InfratoApp / LLM prices / Mistral AI

Mistral AI API pricing

Mistral AI has 10 models with a token price on record: input $0.10 to $16.00 and output $0.10 to $16.00 per million tokens at the standard tier. Each rate links to the page it was read from, most recently on 2026-10-22.

Mistral AI sells models it makes, so these are its own list prices.

Price a workload on these models →Every seller →Mistral AI’s own pricing page →

10

models with a price

10

publish a cached rate

○

batch listings

no batch tier on record

2026-10-22

most recent read

Every model Mistral AI prices

Mistral AI token prices per million tokens, by model and tier, each with the page it was read from
Model	Tier	Input $/M	Output $/M	Cached $/M	Context	Read from
mistralai/codestral-↗	Standard	$0.30	$0.90	$0.03	○	2026-10-22JSON
mistralai/ministral-3-14b	Standard	$0.20	$0.20	$0.02	○	2026-10-22JSON
mistralai/ministral-3-3b	Standard	$0.10	$0.10	$0.01	○	2026-10-22JSON
mistralai/ministral-3-8b	Standard	$0.15	$0.15	$0.015	○	2026-10-22JSON
mistralai/mistral-large-3	Standard	$0.50	$1.50	$0.05	○	2026-10-22JSON
mistralai/mistral-large-4	Standard	$1.36	$4.18	$0.14	○	2026-10-22JSON
mistralai/mistral-medium-3.5	Standard	$1.50	$7.50	$0.15	262K	2026-10-22JSON
mistralai/mistral-small-4	Standard	$0.15	$0.60	$0.015	○	2026-10-22JSON
mistralai/voxtral-tts-↗	Standard	$16.00	$16.00	$1.60	○	2026-10-20JSON
mistralai/z.ai-glm-5.3	Standard	$1.40	$4.40	$0.14	○	2026-10-22JSON

○ means no answer on record, not zero. Rates older than 21 days are hidden rather than shown stale. Input and output are priced separately, so a figure built from input alone is a different, smaller number.

Other sellers

OpenRouter 385 Novita AI 65 Together AI 20 Anthropic 19 Baseten 16 Fireworks AI 13 OpenAI 13 FriendliAI 7 Google DeepMind 7 SambaNova 6 Moonshot AI 4 AI21 Labs 2 DeepSeek 2 Cerebras Inference 1 xAI 1

Frequently asked

How much does the Mistral AI API cost per million tokens?

It depends on the model. At the standard tier, input is $0.10 to $16.00 and output is $0.10 to $16.00 per million tokens across 10 models on record. Input and output are priced separately, and every model's rates are in the table above with the page each was read from.

Does Mistral AI price cached input or a batch tier?

10 models carry a cached input rate on record. No batch tier is on record. A missing rate means no answer on record, not that none exists.

Where do these Mistral AI prices come from?

Each row links to the page it was read from and shows the date it was read. A price older than 21 days is hidden rather than shown stale, and a rate we have not read is left blank rather than estimated.

How these prices are read

Every rate here links to the page it was read from and carries the date it was read. Nothing is averaged or estimated, and a rate we have not read is shown as missing instead of guessed. How any figure is gathered, and what is refused, is on methodology.