Simulating realistic Agentic workflow parameters (40,000 in / 1,500 out with 65% cache reuse). Codestral 2508 (OpenRouter) delivers a 57% cost reduction over GLM 4.7 (Zhipu).
| Traffic Volume Tier | GLM 4.7 (Zhipu) Monthly | Codestral 2508 (OpenRouter) Monthly | Monthly Savings by picking Codestral 2508 (OpenRouter) |
|---|---|---|---|
| 1,000 reqs/mo (Dev/Testing) | $14.56 | $6.33 | Save $8.23 / mo |
| 10,000 reqs/mo (Small App) | $145.60 | $63.30 | Save $82.30 / mo |
| 100,000 reqs/mo (Growth Production) | $1,456.00 | $633.00 | Save $823.00 / mo |
| 1,000,000 reqs/mo (Scale SaaS) | $14,560.00 | $6,330.00 | Save $8,230.00 / mo |
Codestral 2508 (OpenRouter) is 57% cheaper for Agentic workflow workloads. At standard Agentic workflow parameter ratios (40,000 input tokens, 1,500 output tokens, 65% cache hit), Codestral 2508 (OpenRouter) costs $0.00633 per request compared to $0.0146 on GLM 4.7 (Zhipu).
GLM 4.7 (Zhipu) offers a context window of 200,000 tokens (max output: 131,072), while Codestral 2508 (OpenRouter) offers 256,000 tokens (max output: 204,800).
At 100,000 requests per month, using Codestral 2508 (OpenRouter) saves $823.00 every month (or $9,876.00 annually) compared to GLM 4.7 (Zhipu).
Prompt caching is the single biggest lever — each step re-reads prior context. Summarize or prune tool outputs before appending them to the transcript. Cap the step budget; runaway loops are the #1 surprise on agent invoices.