Mistral: Codestral 2508

Codestral 2508 is a specialized language model engineered specifically for software development and technical workflows. By moving away from general-purpose conversational utility, the architecture focuses on a dedicated fill-in-the-middle approach that prioritizes syntax logic and structural precision. This design intent makes the model particularly effective for IDE integrations and high-frequency coding tasks, such as automated test generation and real-time code correction, where maintaining logical consistency is paramount.

The model leverages a refined weight distribution that emphasizes technical performance over conversational filler, resulting in a lean design that balances speed with complex reasoning capabilities. Its capacity to handle large-scale inputs allows developers to process extensive codebases and long-form documentation within a single context window. This focus on specialized utility positions the model as a practical tool for developers who require high-quality code generation and function calling without the overhead of larger, general-purpose flagship models.

Kilo Gatewaymistralai/codestral-2508codestral

Open provider page Provider docs

Quick Info

Provider: Kilo Gateway
Model key: mistralai/codestral-2508
Release date: Aug 1, 2025
Last updated: Aug 1, 2025
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.30
Output token cost: $0.90

Limits

Output tokens: 204,800 tokens
Context window: 256,000 tokens

Transparent token rates

Compare Mistral: Codestral 2508 pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Mistral: Codestral 2508

Kilo Gateway

Coverage

Mistral AI released Codestral 25.08, its flagship open coding model, as reported by AICoder on August 30, 2026. The release is explicitly named Codestral 25.08 and is described as open-source, with weights published on Hugging Face alongside hosted API endpoints and instant Ollama setup support. The report frames it as The release centers on a native 256k token context window that ingests entire repository trees for accurate symbol resolution, paired with enhanced Fill-in-the-Middle (FIM) optimization delivering sub-second multi-line autocomplete across more than 80 programming languages. The article claims Codestral 25.08 outperform

Kilo Gateway

CoverageBenchmark

OpenRouter's model page for Mistral: Codestral 2508 (mistralai/codestral-2508) lists it as a Mistral-released coding model positioned for low-latency, high-frequency tasks including fill-in-the-middle (FIM), code correction, and test generation. The page records a 256K context window, text input/output only, a March 20 The page also surfaces provider-routing metadata for Mistral and Mistral (EU), including weighted-average effective pricing, cache hit rates, tool-call and structured-output error rates, and uptime — content characteristic of a gateway/aggregator listing rather than a creator-authored model announcement. Because the su