Models & Pricing
ZBStream bills input, cache-hit, and output tokens separately based on the model actually routed to. Final prices follow the current console configuration.
Model Details
A single API key can call every model enabled for its application. Model names must exactly match the ids returned by the model list.
| Example model | Best for | Availability |
|---|---|---|
| deepseek-v4-pro | Complex reasoning, coding, and long-form tasks | Depends on current channels |
| deepseek-v4-flash | Low-latency chat and high-frequency calls | Depends on current channels |
How does billing work?
Credits are the sum of cache-miss input tokens, cache-hit tokens, and output tokens, settled after a successful request.
- Cache-miss input credits = cache-miss input tokens × prompt rate.
- Cache-hit credits = cache-hit tokens × cache_hit rate.
- Output credits = output tokens × completion rate.
- Failed requests with no token consumption are not billed as normal usage.
Note
Model rates may combine the upstream price, channel markup, and the application's credits-per-yuan ratio. Always refer to the prices shown in the console.