Codestral 25.08
Mistral's enterprise code model: 80+ languages, low-latency fill-in-the-middle, at the heart of the Mistral coding stack.
About
Codestral 25.08 is Mistral AI's code generation model, announced July 30, 2025 as the centerpiece of a complete enterprise coding stack alongside Devstral (agentic workflows), Codestral Embed (code search) and the Mistral Code IDE extension built on the open-source Continue project. It covers 80+ programming languages with fill-in-the-middle completion tuned for low-latency IDE scenarios, and Mistral positions the stack for cloud, VPC or fully on-prem deployment so regulated industries keep code inside their own boundaries.
Note the licensing shift: the original Codestral 22B shipped under the non-commercial Mistral Non-Production License, and the 25.x revisions are proprietary API models — unlike Mistral's Apache-2.0 small models, there are no public weights for Codestral 25.08. Public validation rests on Mistral's reports of live IDE testing across production codebases rather than any published benchmark table, so capability claims should be read as provider-reported.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
No independent benchmark table verified; Mistral reports validation through live IDE usage across production codebases (provider-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Mistral announcement Jul 30, 2025 with the full coding stack (Codestral 25.08, Devstral, Codestral Embed, Mistral Code); 80+ language coverage; enterprise VPC/on-prem deployment story with named partners (Capgemini, SNCF per press). Closed weights, API-only. Release facts corroborated across Mistral's official news page and 2026 recap coverage. Strengths: coherent enterprise coding platform, European AI-sovereignty positioning, real IDE distribution. Docked for: (1) no published benchmarks — capability claims are provider-reported; (2) proprietary licensing is a step back from the open-weights Codestral 22B era.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 2.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Mistral announcement 2025-07-30; full coding stack (Devstral, Codestral Embed, Mistral Code); 80+ languages; enterprise VPC/on-prem
- · Docked: No published benchmarks — claims provider-reported
- · Docked: Proprietary licensing, a step back from open-weights Codestral 22B