Simulating realistic RAG / search-augmented answers parameters (8,000 in / 500 out with 50% cache reuse). Codestral 2508 (OpenRouter) delivers a 85% cost reduction over Pixtral Large.
| Traffic Volume Tier | Pixtral Large Monthly | Codestral 2508 (OpenRouter) Monthly | Monthly Savings by picking Codestral 2508 (OpenRouter) |
|---|---|---|---|
| 1,000 reqs/mo (Dev/Testing) | $11.80 | $1.77 | Save $10.03 / mo |
| 10,000 reqs/mo (Small App) | $118.00 | $17.70 | Save $100.30 / mo |
| 100,000 reqs/mo (Growth Production) | $1,180.00 | $177.00 | Save $1,003.00 / mo |
| 1,000,000 reqs/mo (Scale SaaS) | $11,800.00 | $1,770.00 | Save $10,030.00 / mo |
Codestral 2508 (OpenRouter) is 85% cheaper for RAG / search-augmented answers workloads. At standard RAG / search-augmented answers parameter ratios (8,000 input tokens, 500 output tokens, 50% cache hit), Codestral 2508 (OpenRouter) costs $0.00177 per request compared to $0.0118 on Pixtral Large.
Pixtral Large offers a context window of 128,000 tokens (max output: 8,192), while Codestral 2508 (OpenRouter) offers 256,000 tokens (max output: 204,800).
At 100,000 requests per month, using Codestral 2508 (OpenRouter) saves $1,003.00 every month (or $12,036.00 annually) compared to Pixtral Large.
retrieved-context dominates cost — tune top-k and chunk size before switching models. Re-rank and drop marginal chunks; halving context roughly halves input cost. Deduplicate repeated chunks across queries to raise the cache hit rate.