Codestral 25.08 vs DeepSeek-V3.2
Price and specs side by side · USD per 1M tokens · verified 2026-11-17
// side by side
| metric | Codestral 25.08 | DeepSeek-V3.2 |
|---|
| input / 1M | $0.30 | $0.26 |
| output / 1M | $0.90 | $0.38 |
| blended (3:1) | $0.45 | $0.29 |
| cache read / 1M | $0.03 | $0.13 |
| context | 256K | 163.8K |
| max output | 8.2K | 64K |
| input types | text · pdf | text |
| reasoning | no | yes |
| tool calling | yes | yes |
| structured output | yes | yes |
| open weights | no | yes |
| released | 2025-07-30 | 2025-12-01 |
| knowledge cutoff | n/a | 2024-07 |
| lab | Mistral | DeepSeek |
// monthly cost scenarios
| workload | tokens in / out | Codestral 25.08 | DeepSeek-V3.2 |
|---|
| Chatbot | 10M / 2M | $4.80 | $3.36 |
| RAG app | 50M / 5M | $19.50 | $14.90 |
| Coding agent | 20M / 10M | $15.00 | $9.00 |
// verdict
- DeepSeek-V3.2 is 35.6% cheaper than Codestral 25.08 on a blended 3:1 input/output basis.
- Codestral 25.08 has the larger context window (256K vs 163.8K tokens).
- Only DeepSeek-V3.2 supports built-in reasoning.
- Chatbot (10M input / 2M output tokens per month): DeepSeek-V3.2 costs $3.36 vs $4.80 for Codestral 25.08, saving $1.44 per month.
- RAG app (50M input / 5M output tokens per month): DeepSeek-V3.2 costs $14.90 vs $19.50 for Codestral 25.08, saving $4.60 per month.
- Coding agent (20M input / 10M output tokens per month): DeepSeek-V3.2 costs $9.00 vs $15.00 for Codestral 25.08, saving $6.00 per month.
// Codestral 25.08 or DeepSeek-V3.2: which one to choose
- Choose DeepSeek-V3.2 for cost-sensitive, high-volume workloads.
- Choose Codestral 25.08 for long documents and large codebases (256K-token context).
- Choose DeepSeek-V3.2 if you need built-in reasoning.
- Choose DeepSeek-V3.2 if you want open weights you can self-host.