Mistral
Efficient

Codestral Mamba

Code Mamba

Mamba-architecture variant of Codestral, linear-time inference at very long context, Apache 2.0.

Codestral Mamba is a efficient AI model from Mistral. It costs $0.2500 per million input tokens and $0.2500 per million output tokens (blended $0.2500/M), with a 256,000-token context window.

RatesUSD per million tokens. Blend is 70% input, 30% output, our calculation.

INPUT
$0.2500/M
per million input tokens
OUTPUT
$0.2500/M
per million output tokens
BLENDED 70/30
$0.2500/M
unchanged since 3 May
CONTEXT
256,000
tokens

Hand-maintained from Mistral's published pricing; no check date on record.

What it is good at
  • Mamba (SSM) architecture
  • Linear-time long context
  • Open weights
Typical use cases
  • Repo-scale code analysis
  • Long-document architectural research

Benchmarks

vs. best public score
Python function synthesis from docstrings. Source: Mistral AI (self-reported), HumanEval, read 2026-11-10.
Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources, treat as directional rather than authoritative.

More from Mistral

See all 15 →

Frequently asked questions

How much does Codestral Mamba cost?

Codestral Mamba costs $0.2500 per million input tokens and $0.2500 per million output tokens, for a blended reference rate of $0.2500 per million tokens.

What is Codestral Mamba's context window?

Codestral Mamba supports up to 256,000 tokens of context in a single request.

What is Codestral Mamba best for?

Codestral Mamba is well suited to Mamba (SSM) architecture, Linear-time long context and Open weights.

Who makes Codestral Mamba?

Codestral Mamba is developed and served by Mistral. It was released in Jul 2024.

Terms used on this page