Multi-model orchestration, delivered as one API. Intelligent routing across a pool of the strongest open models.
Replies
Best
Finally a way to stop juggling five different API keys. The routing just picks the right model for my prompts without me thinking about it, and the latency has been solid on everything ive thrown at it.
Report
Honestly the routing logic here feels really thought out, like picking the best model per call without me having to babysit it is kind of magical. Solid execution from the team.
Report
Routing across a pool of frontier models is a crowded but real problem - how are you handling the API surface fragmentation, e.g. do you expose an Anthropic-style /v1/messages alongside OpenAI-style chat/completions, or normalize everything to one schema? Curious too whether routing decisions are latency/cost-based or capability-based, since that shapes how coding agents like Claude Code or Cursor behave against it.
Replies
Finally a way to stop juggling five different API keys. The routing just picks the right model for my prompts without me thinking about it, and the latency has been solid on everything ive thrown at it.
Honestly the routing logic here feels really thought out, like picking the best model per call without me having to babysit it is kind of magical. Solid execution from the team.
Routing across a pool of frontier models is a crowded but real problem - how are you handling the API surface fragmentation, e.g. do you expose an Anthropic-style /v1/messages alongside OpenAI-style chat/completions, or normalize everything to one schema? Curious too whether routing decisions are latency/cost-based or capability-based, since that shapes how coding agents like Claude Code or Cursor behave against it.
Pazi
Congrats on launching AI Gateway! Ai Gateway is a genuinely exciting direction. Cheering you on and wishing you a fantastic launch! 🚀