DeepSeek API pricing
USD per 1M tokens. Crossed-out amounts are reference list rates; green amounts include the 15% discount. These are public display names, not API IDs.
| Model | Input | Cached input | Output |
|---|---|---|---|
| DeepSeek V4 Pro | |||
| DeepSeek V4.1 Flash |
Reference list: September 29, 2026. Same rates as the Models page; confirm live account billing before use.
Estimate your workload cost →Start with the task, then test the model.
Evaluate DeepSeek with tasks drawn from your own workflow. Compare output quality, latency, and actual usage cost before selecting a production model.
- Code explanation and review
- Structured problem-solving
- Multi-step business analysis
Retrieve the exact API model ID
Use your server-side key to list the models available to your account. Copy an ID exactly as returned; do not derive it from the display name above.
curl https://api.sinorouter.ai/v1/models \
-H "Authorization: Bearer $SINOROUTER_API_KEY"Use that ID with /v1/chat/completions for supported text models. Follow the text API guide for cURL, Python, and JavaScript examples.
What to confirm before production
- Check whether reasoning is controlled by the model ID or a separate parameter on your route.
- Test tool use and reasoning responses independently. Client support alone is not enough.
- Measure input, cached input, and output usage separately; cache hits are not guaranteed.
- Confirm geographic availability, upstream data handling, and account limits.
Read the production checklist before increasing traffic.
From a test to your application
Use the same representative prompts and output limits across tests. Compare the cost of a completed task, including repeated calls, tool loops, and actual cached usage.
Provider reference
DeepSeek API documentation ↗Provider documentation explains its own service. It does not establish that every feature, region, or rate is available on your SinoRouter route. Confirm route-specific support before use.
Editorial update: October 4, 2026 · SinoRouter