mistral-large-4 price up to 50% cut on 2026-10-29

On 2026-10-29, Mistral changed the published API price of mistral-large-4. The superseded price had stood for 5 days. This is recorded change 2 of 2 for this model in the index.

Every metered token direction on mistral-large-4 got cheaper that day. A cut is the direction people notice least, because a lower bill rarely prompts anyone to check what changed.

What changed

Per 1M tokensBeforeAfterChange
Input$1.36$0.70-48.5%
Output$4.18$2.10-49.8%
Cache read$0.14$0.070-50%

What it meant in practice

On a reference workload of 50M input and 5M output tokens a month, this change moved the bill from $88.90 to $45.50 a month, a difference of $43.40. The workload is a fixed reference used on every event page so the figures are comparable between them, not a measurement of anyone's usage.

Your recorded history should not change when prices do

A cost tracker carrying a single current-price table would reprice every mistral-large-4 session you ran before 2026-10-29 at the new rate, silently restating months you already paid for. See what this change does to a real session, and why dated windows are the fix.

Provenance

Every value here carries a first-party source, and superseded prices are kept rather than overwritten. Windows before the index began are reconstructed from web archives and labelled as archived rather than claimed as first-party.

WindowSource
Input, 2026-10-24 to 2026-10-29https://docs.mistral.ai/inference/pricing
Input, from 2026-10-29https://docs.mistral.ai/inference/pricing
Output, 2026-10-24 to 2026-10-29https://docs.mistral.ai/inference/pricing
Output, from 2026-10-29https://docs.mistral.ai/inference/pricing
Cache read, 2026-10-24 to 2026-10-29https://docs.mistral.ai/inference/pricing
Cache read, from 2026-10-29https://docs.mistral.ai/inference/pricing

Related

Derived from the AI Price Index, last updated 2026-10-30. Licensed CC BY 4.0: reuse this with credit to the dataset page.