LLM API prices / Mistral Large 4

Mistral Large 4 API pricing

Last checked: 2026-10-07. Mistral's official model page lists Mistral Large 4 as a public preview. Its displayed discounted rate is $0.68 per million input tokens, $2.09 per million output tokens and $0.07 per million cached input tokens. The page also shows struck-through higher figures, so these rates should be rechecked before a long-running budget is approved.

When this model fits

It can be a candidate for applications that need one general-purpose API for text and image understanding. The published context window is 1 million tokens, but a large window does not mean every request should fill it. Retrieval, prompt trimming and smaller models can reduce cost when most tasks use only a few thousand tokens. Preview availability and supported regions can change; inspect the model card and account console for production access.

A transparent cost example

Suppose a service sends 1,000 input tokens and receives 500 output tokens, 1,000 times daily for 30 days. Input volume is 30 million tokens and output volume is 15 million. With no cache hits, Large 4 costs 30 × $0.68 + 15 × $2.09 = $51.75. At the same token counts, GPT-6.1 Sol costs 30 × $2 + 15 × $10 = $210.00, based on OpenAI's official standard rates. These totals compare token charges only; they make no claim about answer quality or equivalent output lengths.

Caching changes the first total only when requests meet Mistral's cache conditions. If 10 million of those input tokens were cache reads, their token charge would be 10 × $0.07 instead of 10 × $0.68; cache writes and other features need their own review. Do not apply a cache discount to all traffic without measured cache hits. Check billing details for batch and any region-specific price.