Inference pricing
Understand included inference credits and the token charges that apply after those credits are used.
Every Jotsu plan includes a fixed amount of inference credit. After that credit is used, continued inference is charged at the underlying model cost plus 15%.
Included credits
Inference credits cover model requests routed through Jotsu. This includes direct API calls, AI features in your applications, and interactive Codex sessions using Jotsu inference.
The amount of included credit and its reset date depend on your plan. Unused credits do not become cash and do not offset future plan periods.
Usage after included credits
Jotsu calculates each request from the model's input and output token prices. Once the included credit is exhausted, the charge is:
underlying model cost × 1.15
For example, a request with an underlying model cost of $1.00 is charged at $1.15 after included credits are used.
Input, cached input, and output tokens may have different prices. Long-context requests may also use a different price tier.
Control overage
Your organization's overage budget limits additional inference spending. Jotsu stops new charged requests when that budget is exhausted.
Plans without paid overage stop inference when their included credit reaches zero.
Codex with connected credentials
Interactive Codex requests made through credentials connected by your organization do not consume Jotsu inference credits. The connected provider or subscription controls their cost and limits.
Application inference and other Jotsu-managed model calls continue to use Jotsu credits even when interactive Codex uses connected credentials.
See Codex model access to compare the two access methods.