Changelog

Added the full GPT-5.6 family (Sol, Terra, Luna) and Gemini 3.1 Pro & 3.5 Flash to the gateway. Every new frontier model ships at about a third of provider list price.

Requests now route to the fastest healthy provider automatically, with instant failover when an upstream degrades — no client-side changes required.

Repeated context is cached across calls, cutting cached-input costs by up to 90% on supported models.

Real-time usage analytics, per-key rate limits, and hard budget caps so teams never get a surprise bill.

Launched the OpenAI-compatible gateway — call Claude, GPT, Gemini, and Grok through a single endpoint and one API key, at a third of the price.