← Back to models
↗↗↗
Mistral
Codestral 25.08
Mistral’s specialist coding model for low-latency code generation and fill-in-the-middle workflows.
Overall score
87
codingfill-in-the-middlelow-latency
Context window
128K tokens
Speed
Fast
Input pricing
$0.30 / 1M input tokens
Output pricing
$0.90 / 1M output tokens
Score breakdown
Capability88
Use-case fit92
Cost efficiency90
Speed93
Reliability84
Agent readiness89
Ecosystem82
Scores combine benchmark signals, product experience, and editorial weighting. Use them as a practical guide, not an absolute truth claim.
Best for
CodingAgent automation
Works with
Mistral APIcode completionFIM editors
Modalities
text
Sources & trust
Officially verified core fields
Official linkSummaryDescriptionContext windowPricingPricing pageTool support
Editorial fields such as shortlist guidance, strengths, caveats, and scoring remain clearly separated from official provider data.
Mistral official
Official site · Tier 5 · Apr 13, 2026
Official link
Codestral 25.08 docs
Official docs · Tier 5 · Apr 13, 2026
SummaryDescriptionContext windowPricingPricing pageTool support
Codestral 25.08 VerdictLens review
Manual review · Tier 3 · Apr 9, 2026
Best-fit guidanceWorks-with guidanceStrengthsCaveatsOverall scoreScore breakdown
Last verified: Apr 13, 2026
Strengths
- Adds a purpose-built coding specialist instead of only general Mistral models.
- Official docs clearly call out low-latency code generation and FIM workflows.
Things to watch
- This is a narrower coding-focused model, not the best all-purpose research tier.
- Teams still need their own evals for repo-specific coding behavior.