GPT-5.6 Sol runs 50% off on Vercel's AI Gateway until Sept 18
Half price on every request, no code change — for a month, OpenAI's up-to-14x model is cheap to stream. Plus: Cursor Origin repos now auto-deploy to Vercel.
Copy markdown
Half off, every request, no code change
From Aug 17 through Sep 18, every call you route to GPT-5.6 Sol through Vercel's AI Gateway is 50% off — same model ID, no SDK swap. If Sol is already in your stack, your inference bill just halved for a month.
And Sol is the fast one
Sol is OpenAI's ultrafast tier — the preview clocks up to 14x normal speed (about 750 tokens/sec on Cerebras). Cutting its price in half makes real-time token streaming the cheapest it has been.
Build this weekend
Half-price, sub-second output turns things you'd normally rate-limit into things you leave running: live voice loops, agents that stream their reasoning, type-ahead coding UIs. Prototype now while the meter's cheap.
Also new on Vercel: Cursor Origin auto-deploys
The day-old Origin code host — Cursor's GitHub alternative — now wires straight into Vercel: connect an Origin repo and it ships Preview and Production deploys automatically, so agent-written branches deploy without leaving Cursor.