Vercel's AI Gateway adds OpenAI's Ultrafast tier for GPT-6 Astra
Vercel's AI Gateway now supports OpenAI's Ultrafast service tier for GPT-6 Astra, which is meant to deliver faster output for interactive applications and rapid coding iterations. Requests served at Ultrafast are billed at 6× the standard per-token rate.
- AI Gateway now supports OpenAI's Ultrafast service tier for GPT-6 Astra.
- Ultrafast can be requested via the AI SDK, Chat Completions API or Responses API for openai/gpt-6-astra.
- Ultrafast requests are billed at 6× the standard per-token rate.
- The tier supports US and global processing; requests pinned to regions such as the EU run at the standard tier.
- Standard processing remains the default when no service tier is specified.
How to request Ultrafast
Developers can request the tier for the model openai/gpt-6-astra through the AI SDK, the Chat Completions API or the Responses API. In the AI SDK example published by Vercel, the call to generateText sets the provider option serviceTier to ultrafast under the openai key. The example also restricts routing to OpenAI with the gateway option only: ['openai'].
The sample prompt in the changelog is "Investigate the failing tests and propose a fix.", which reflects the coding-iteration use case Vercel names alongside interactive applications.
Guidance for tool-heavy workflows
For workflows with frequent tool calls, OpenAI recommends using the Responses API over a persistent WebSocket connection. According to the source, this reduces overhead between turns. Vercel points to its Ultrafast service-tier examples for persistent connections, including the AI SDK over WebSocket.
Regions, defaults and billing
Ultrafast supports US and global processing. Requests pinned to unsupported regions, such as the EU, run at the standard (default) tier instead. Standard processing stays the default when no service tier is specified, so existing integrations are not switched over automatically.
Pricing is the main trade-off. Requests served at Ultrafast are billed at 6× the standard per-token rate. Requests that fall back to another tier are billed at the rate for the tier actually served. Vercel directs users to the GPT-6 Astra model page for current rates.
What it means for developers
The change gives teams using AI Gateway an opt-in way to trade higher cost for faster output on GPT-6 Astra. It fits latency-sensitive uses such as interactive apps and fast coding loops, while other workloads can stay on the standard tier. Teams with EU-pinned requests should note that those calls will run at the standard tier and be billed accordingly.
Vercel: OpenAI Ultrafast mode now available on AI Gateway
