Codestral 2508

Provider: Mistral AI EU Official API price

Input price
$0.30
/ 1M tokens
Output price
$0.90
/ 1M tokens
Cache price
$0.03
/ 1M tokens (read)
Context
256K tokens
Max output 205K tokens

Overview

Mistral AI's Codestral 2508 is a specialized model designed for code-centric workloads. It features a very large context window and specific structured output capabilities. This model operates in a dedicated reasoning specialist tier for developers and applications requiring extensive code understanding.

Strengths

  • Provides an impressively large 256,000 token context window for extensive code analysis.
  • Supports function-calling for robust integration with external tools and APIs.
  • Offers JSON mode for reliable generation of structured data outputs.
  • Features a unique cache read price of $0.03/1M tokens for optimized re-processing.
  • Its output price is more competitive than some similarly-priced peers like Gemini 2.5 Flash.

Best use cases

  • Analyzing, refactoring, or generating code for large and complex projects.
  • Building intelligent agents capable of interacting with APIs via function calls.
  • Extracting structured information from codebases or documentation.
  • Automating code review or generating test cases within vast repositories.
  • Developing applications that require highly accurate, structured code outputs.

Pricing position

Codestral 2508 is positioned as Mistral AI's highest-priced model, with input at $0.3/1M tokens and output at $0.9/1M tokens. This makes it significantly more expensive than Mistral Nemo ($0.02/1M in, $0.03/1M out) or Mistral Small 3 ($0.05/1M in, $0.08/1M out). When compared to other models with a $0.3/1M input price, its output cost of $0.9/1M tokens is lower than Gemini 2.5 Flash ($2.5/1M) and Qwen3.5 Plus ($1.8/1M), but higher than Grok 3 Mini ($0.5/1M) and comparable to GLM 4.6V ($0.9/1M).

Caveats

The maximum output token limit for Codestral 2508 is not specified, which could be a consideration for tasks requiring very long generations. Users seeking a general-purpose model or a more cost-optimized solution for simpler tasks might find other Mistral models, such as Mistral Nemo, more suitable. Additionally, there is no free tier available for this model.

Notes drafted by an LLM from the figures on this page (2026-05-21) and rewritten when the price changes.

Key facts

Price basis
Official API price
Max output
205K tokens
Knowledge cutoff
2025-03-31
Listed on OpenRouter
2025-08-01
Tracked since
2026-05-08
Last updated
2026-11-10

Features

Function callingJSON output

Monthly cost estimate

Scale Input (M) Output (M) Monthly
0.1M (hobby) 0.1 0.05 $0.08
1M (small app) 1 0.5 $0.75
10M (mid scale) 10 5 $7.50
100M (large) 100 50 $75.00

※ Cache and batch discounts not factored in. Detailed calculator

Official resources

Other models from Mistral AI

Similar-priced models

Source: OpenRouter API (mistralai/codestral-2508)