OpenRouter model

Mistral: Codestral 2508

Mistral's cutting-edge language model for coding released end of July 2025. Codestral specializes in low-latency, high-frequency tasks such as fill-in-the-middle (FIM), code correction and test generation. [Blog Post](provider documentation

Provider
OpenRouter
Availability
Available
Model ID
mistralai/codestral-2508
Snapshot
2026-10-07
Evaluation Verdict

Mistral Codestral 2508 is a specialized autonomous code generation and Fill-In-the-Middle (FIM) model (43.5% SWE-bench Verified, 90 TPS), optimized for high-speed file editing and syntax completion.

Mistral: Codestral 2508 Benchmarks & Agentic Accuracy

Verified primary research & developer preference metrics

Primary Sources Verified
SWE-bench Verified
43.5%
FIM / AgenticMistral AI System Card ↗
LiveCodeBench
42%
StandardLiveCodeBench ↗
LMSYS Coding Elo
1305Elo
Coding ArenaLMSYS Arena ↗
AIME 2024 / 2025
38%
StandardMistral AI System Card ↗
Pricing ($ / 1M tokens)
$0.3/ $0.9
$0.08 cached readStandard API Pricing

What This Means in Real Software Development

  • Specialized for code generation with 256k context window and native Fill-In-the-Middle (FIM) support.
  • Scores 43.5% on SWE-bench Verified with rapid single-pass patch generation.
  • High throughput (90 TPS) provides fast turn-around in interactive coding sessions.
  • Economical $0.30/M input pricing on Mistral AI and OpenRouter.

Swarm Multi-Agent Topology

Tier 2: Specialized Code Synthesis EngineRole: Coder Subagent

Best Paired With: Claude 3.7 Sonnet as Lead Planner

Excellent for isolated function implementation, refactoring routines, and syntax patching in Task Programs.

Git Worktree Isolation Advantage:

Authors and validates code diffs within isolated Git worktrees.

What Developers & Practitioners Are Saying

Real-world field observations, production strengths, and developer consensus

Field Consensus

"Mistral's dedicated code generation model. Extremely fast with a focused 256k context window, specifically tuned for fill-in-the-middle (FIM) and code completion."

Verified Real-World Strengths

  • Superior fill-in-the-middle (FIM) performance for IDE tab completion
  • Low latency (~90 TPS) with affordable pricing ($0.30/M in, $0.90/M out)
  • Fluent in over 80 programming languages

Production Caveats & Gaps

  • SWE-bench score (34.0%) reflects weaker high-level multi-file planning
  • Better suited as an implementation subagent than a lead architect
Benchmark Reality Check

What the benchmark tests: Focuses on code completion, syntax prediction, and isolated function synthesis.

Production caveat: Do not use as a primary planner; deploy strictly for localized function implementation.

Primary Sources & Methodology Citations

Mistral: Codestral 2508 model overview

Mistral: Codestral 2508 is listed in the public Swarm catalog through OpenRouter. This page summarizes the verified catalog fields used by Swarm without claiming real-time access for every provider account.

Provider
OpenRouter
Family
OpenRouter
Context window
256K tokens
Thinking support
Not supported
Availability status
Available
Catalog ID
openrouter/mistralai/codestral-2508
  • File input
  • Text input
  • Text output
  • Tools
  • Structured output
  • JSON

Run Mistral: Codestral 2508 in Swarm

Verified CLI workflow & multi-agent architecture

Quickstart Guide →
bash — swarm run
$swarm ctl model set --provider openrouter --model openrouter/mistralai/codestral-2508

# Or switch interactively inside the terminal TUI:

>/models

Subagent Role Matrix

Assign Mistral: Codestral 2508 to specialized system roles in ~/.swarm/config.json:

  • Primary / Planner: Orchestrates full task DAGs and multi-stage pipelines.
  • Coder Subagent: Generates code and unit tests inside private Git worktrees.
  • Finder Subagent: Performs read-only code search, symbol discovery, and mapping.
← All OpenRouter models

Public catalog snapshot. Catalog availability is not a real-time provider status or an access guarantee.