Ask

Codestral vs Mistral Small 3

Detailed pricing comparison and cost analysis.

Updated August 2026

Cost Simulator

Codestral Cost
$0.48
Mistral Small 3 Cost
$0.05
Mistral Small 3 is 89% cheaper
FeatureCodestralMistral Small 3
ProviderMistralMistral
Input Price (1M)$0.30$0.03
Output Price (1M)$0.90$0.11
Context Window256,00033,000

Verdict

Codestral costs $0.30 per 1M input tokens and $0.90 per 1M output tokens. Mistral Small 3 costs $0.03 per 1M input tokens and $0.11 per 1M output tokens. Mistral Small 3 is 90% cheaper on input tokens than Codestral. For output tokens, Mistral Small 3 is the more affordable option at $0.11/1M vs $0.90.

On context window, Codestral supports 256,000 tokens — meaning it can fit more conversation history, documents, or code in a single request. This matters for RAG pipelines, long document analysis, and agentic workflows where context builds up over many turns.

When to choose Codestral

  • ✓ You need a larger context window (256,000 tokens)
  • ✓ You are already integrated with Mistral

When to choose Mistral Small 3

  • ✓ You need the lowest input token cost ($ 0.03/1M)
  • ✓ Your workload is output-heavy — Mistral Small 3 generates text cheaper
  • ✓ You are already integrated with Mistral

Use the calculator above to simulate your specific workload and find the exact break-even point. For most applications, the cheapest model is the one that minimises your total monthly bill given your input-to-output token ratio.

Frequently Asked Questions

Is Codestral cheaper than Mistral Small 3?

Mistral Small 3 is cheaper on input tokens at $0.03/1M vs $0.30/1M for Codestral — a 90% saving.

What is the context window of Codestral vs Mistral Small 3?

Codestral has a 256,000-token context window. Mistral Small 3 has a 33,000-token context window. Codestral supports the larger context, suitable for longer documents and agentic workflows.

Which model is better: Codestral or Mistral Small 3?

The best choice depends on your use case. For cost efficiency on input tokens, Mistral Small 3 is the cheaper option. For maximum context length, Codestral supports 256,000 tokens. Use the comparison table above to find the right fit for your workload.