Google Antigravity lists Claude Opus/Sonnet 5.5: paid non-trial Pro + Ultra only; Claude 4.6 and GPT-OSS-120b retire Nov 2

Background

Model pickers inside agentic IDEs are becoming procurement surfaces: teams want Gemini quotas and Claude’s latest coding feel without three invoices. Google Antigravity already hosted third-party models, but its docs table lingered on Claude 4.6 after Anthropic shipped Opus 5.5 and Sonnet 5.5 in late September. As of 2026-10-03, developer reports match the live official Models table: Claude Sonnet 5.5 (thinking) and Claude Opus 5.5 (thinking) are listed. Verify at https://antigravity.google/docs/models/ — not an X rumor.

Architecture

Antigravity separates the selectable reasoning model from fixed helper models (e.g. Nano Banana 2 for generative images). Claude 5.5 appears as thinking variants beside Gemini 3.x, Claude 4.6, and GPT-OSS-120b. Plan matrix: Free & Google AI Plus ❌; Google AI Pro ✅ for non-trial subscriptions only; Ultra ✅; Enterprise ❌. Footnote: Claude 4.6 and GPT-OSS-120b remove on 2026-11-02. Mid-turn model switches apply only after the current turn finishes or is cancelled. Plans docs: five-hour + weekly quotas on Pro/Ultra; optional AI Credit overages; no BYOK.

Benchmarks

Antigravity published no new SWE/Terminal scores for this listing—do not invent them. Anchor facts: (1) the docs checkmarks and plan gates; (2) Anthropic’s own Opus/Sonnet 5.5 launch cards (separate from Antigravity A/B). The product bet is same-invoice session-level comparison vs Gemini 3.1 Pro. Enterprise still has no Claude 5.5 checkmark.

Hands-on

Confirm Pro (non-trial) or Ultra on the Models matrix; pick Claude 5.5 under the prompt; if missing, check trial/Free/Plus/Enterprise; set AI Credit Overages to Never; migrate off Claude 4.6 / GPT-OSS-120b before 2026-11-02. Android Studio Rabbit 2 Canary BYOA (ACP) is a different surface from Claude 5.5 inside Antigravity itself.