What Codestral costs for real workloads
| Workload | Estimated cost |
|---|---|
| Support chatbot: 100K messages | $51.00 |
| RAG search: 10K queries with 6K-token context | $22.50 |
| Batch summarization: 1,000 long documents | $6.90 |
| Coding agent: 100 tasks, 50K context x 20 turns | $6.15 |
Chatbot: 800 input + 300 output tokens per message. RAG: 6,000 + 500 per query. Summarization: 20,000 + 1,000 per document. Coding agent: 20 turns per task over a 50K-token context, with 95% of input served from cache where the model supports it.
Codestral vs similarly priced models
| Model | Provider | Input / 1M | Output / 1M | Cached / 1M | Blended |
|---|---|---|---|---|---|
| deepseek-flash | DeepSeek | $0.15 | $0.60 | $0.003 | $0.26 |
| GPT OSS 120B | Groq | $0.15 | $0.60 | — | $0.26 |
| Mistral Small 4 | Mistral AI | $0.15 | $0.60 | $0.015 | $0.26 |
| Qwen3 32B | Groq | $0.29 | $0.59 | — | $0.36 |
| Codestral | Mistral AI | $0.30 | $0.90 | $0.030 | $0.45 |
| DeepSeek V4.1 Flash 1017 | Mistral AI | $0.30 | $1.20 | $0.007 | $0.52 |
| Qwen3.7-Plus | Together AI | $0.32 | $1.28 | — | $0.56 |
| Gemini 3.1 Flash-Lite | $0.25 | $1.50 | $0.025 | $0.56 | |
| Llama 3.3 70B Versatile | Groq | $0.59 | $0.79 | — | $0.64 |
Other Mistral AI models
Codestral pricing FAQ
How much does the Codestral API cost?
Codestral costs $0.30 per million input tokens and $0.90 per million output tokens on Mistral AI, with cached input at $0.030 per million. Prices as of November 8, 2026.
How much does Codestral cost for 1,000 requests?
At 1,000 input and 500 output tokens per request, 1,000 requests cost about $0.75.
Does Codestral have a prompt caching discount?
Yes. Cached input is $0.030 per million tokens, 10x cheaper than fresh input. Workloads that resend the same context, like agents and chat with long system prompts, pay mostly the cached rate.
Is Codestral expensive compared to other LLM APIs?
It ranks #19 cheapest of 76 models we track on a blended price (3 input tokens per output token) of $0.45 per million tokens.
Picking a model for production?
Describe your workload and the Planner estimates monthly token spend across every model we track.