Currently listed through these providers:
Model details
Codestral 2508
Codestral 2508 is a specialized language model built for software development workflows that demand speed and reliability. The model excels at fill-in-the-middle completions, enabling developers to insert code into the middle of existing files without disrupting surrounding context. It handles rapid iterative tasks including bug fixes, test generation, and code search across both single and multi-repository environments, allowing developers to locate internal logic, validation routines, or domain-specific utilities using natural language queries. This positions it as a practical tool for environments where developers need to navigate and modify large codebases quickly.
The 2508 release marks significant improvements over earlier versions, delivering a 30% increase in accepted completions, 10% more retained code after suggestions, and 50% fewer runaway generations—each metric directly improving developer confidence when making large edits. Performance gains extend to academic benchmarks for both short and long-context FIM completion, reflecting the model's ability to maintain accuracy across varying code lengths. As an open-weights model with function calling and tool choice support, it integrates well into development pipelines, IDE workflows, and automated code review systems where precise, controlled generation matters more than raw conversational ability.
Quick Info
Powered by- Provider
- Cortecs
- Model key
- codestral-2508
- Release date
- Jul 30, 2025
- Last updated
- Jul 30, 2025
- Knowledge cutoff
- 2025-03
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $0.334
- Output token cost
- $1.003
Limits
- Output tokens
- 256,000 tokens
- Context window
- 256,000 tokens