Open Brain
VercelEngineering PostOfficial Source

GPT-6.2 Astra and Sol now available on AI Gateway

Vercel AI Gateway exposes GPT-6.2 Astra and Sol to coding agents with model IDs and Responses API support. Astra targets complex codebase work; Sol targets everyday workflows, with Ultrafast available for lower latency at higher per-token rates.

Vercel · Nov 17, 2026
Open Source Open MarkdownOpen JSON
Source Summary

**GPT-6.2 Astra** and **GPT-6.2 Sol** are available on AI Gateway. Astra is aimed at complex, multi-step projects and existing-codebase changes; Sol at everyday coding and agent workflows. Both support the Responses API and Ultrafast service tier. For Codex, Cursor, or another supported coding agent, configure the latest Vercel CLI and select `openai/gpt-6.2-astra` or `openai/gpt-6.2-sol`. Use Ultrafast when latency matters, then compare its per-token cost against your normal tier. The source gives no task-level comparisons for these models on AI Gateway. Its 60% lower Ultrafast output-token price is relative to GPT-6 Astra, not a general cost reduction against standard service.

Practical Implication

No separate implication has been recorded yet.

Agent-Ready Context
**GPT-6.2 Astra** and **GPT-6.2 Sol** are available on AI Gateway. Astra is aimed at complex, multi-step projects and existing-codebase changes; Sol at everyday coding and agent workflows. Both support the Responses API and Ultrafast service tier.
For Codex, Cursor, or another supported coding agent, configure the latest Vercel CLI and select `openai/gpt-6.2-astra` or `openai/gpt-6.2-sol`. Use Ultrafast when latency matters, then compare its per-token cost against your normal tier.
The source gives no task-level comparisons for these models on AI Gateway. Its 60% lower Ultrafast output-token price is relative to GPT-6 Astra, not a general cost reduction against standard service.
Connected Context · Feed7 Judgment

This makes GPT-6.2 Astra and Sol available to supported coding agents through AI Gateway, with distinct stated roles for complex codebase work and everyday workflows. The listing gives a concrete model-selection path but no task-level comparisons; Ultrafast’s price claim is relative to GPT-6 Astra, so it does not establish a general cost saving.

Context Map
toolscoding#coding-agents#model-selection#agent-sdks