Paid-plan Quasar Max chat now runs on Claude Sonnet 5 with a 1M-token context window — same tier picker, stronger tool-heavy editing, and PDF attachments unchanged.
Select Quasar Max in the editor AI composer on a paid plan and chat now routes to Claude Sonnet 5 — near-frontier quality on multi-step site edits, with a 1M-token context window for long threads and PDF attachments.
What we shipped
Quasar Max — the paid-tier chat model in the site editor — now uses Claude Sonnet 5 via the Vercel AI Gateway. The previous Sonnet 4.6 gateway id stays in the legacy map, so stored model preferences and older sessions keep resolving to Quasar Max without a manual reset.
What changed
Model — anthropic/claude-sonnet-5 replaces Sonnet 4.6 for Quasar Max chat
Context window — 1M tokens (up from 200k) for long editing sessions and large site manifests
Tier names — unchanged: Quasar Lite (Seed 1.8) on all plans, Quasar Max on paid plans
Images — Quasar Max image generation still uses GPT Image 2; only the chat backbone changed
What stayed the same
Same composer picker — no new setting to learn
PDF attachments remain Quasar Max only
Tool-based edits, change cards, and apply/revert flow are identical
Quasar Lite is untouched
Why it matters
Site editing chat is tool-heavy: create page, update metadata, reorder folders, adjust styles. Sonnet 5 is tuned for that kind of multi-step work — fewer half-finished runs on complex requests and better follow-through when you ask for several related changes in one turn.The larger context window helps when you keep one thread open across a full site pass — homepage, pricing, docs stubs — without starting over.
Try it
Open a site on a paid plan and click the Quasar tab in the editor rail.
Open the model menu in the composer and select Quasar Max.
Ask for a multi-step edit — for example, "Create /changelog.mdx in Updates, add a hero, and set the summary for SEO."
Attach a PDF reference if you need spec or brand copy in context — Max-only, as before.
