Feature
Low latency
Responses stream token by token as the model writes them.
Requests go straight to the model provider with a thin authentication and billing layer in between. Streaming tokens are passed through as they arrive — no buffering, no batching — so your users see answers immediately. Live service latency is published on the Status page, and every request in your logs shows its own latency.
Streaming
Token by token
Status
Live latency page
Routing
Direct to provider
What you get
- No prompt buffering before streaming
- Live service latency on the Status page
- Every request shows its latency in your logs
- Same streaming behaviour as calling the provider directly
Common use cases
Interactive chatVoice appsCoding assistantsLive support
How it works
- 1Create your accountSign up for TokenRex — no card required, no waitlist.
- 2Top up your walletAdd funds through Razorpay. The confirmed amount is credited to your balance.
- 3Create an API keyGenerate a scoped key from the dashboard. Keys are shown once and stored hashed.
- 4Make your first callPoint your SDK at the TokenRex base URL. Every request is metered against your wallet in real time.
Frequently asked questions
Does TokenRex add latency?
The gateway adds only a thin authentication and billing layer; tokens stream through as the provider emits them.
Where can I see current latency?
The Status page publishes live service latency, and each request in your dashboard logs shows its own timing.