Antigravity's free tier drops to one model today as Claude and GPT-OSS leave
Google's Antigravity docs list three third-party models for removal on 18 November 2026: Claude Sonnet 4.6 (thinking), Claude Opus 4.6 (thinking) and GPT-OSS-120b. The notice is a single footnote on the model availability table, with no blog post, no changelog entry and no recommended replacement, and the docs page carries no last-updated stamp. The removal is not scoped to a plan: the Claude 4.6 pair goes from Free and Google AI Plus and from Pro at $19.99 a month, and GPT-OSS-120b goes from Free, Plus, Pro and Ultra at $99.99. Free and Plus users are left with one model, Gemini 3.8 Flash, while Claude Sonnet 5.5 and Opus 5.5 stay on Pro, Ultra and Enterprise. Source: https://pick-right.com/news/google-antigravity-removes-claude-4-6-gpt-oss-third-party-models-2026-11-02/
Why it matters
This is dated today and it directly affects anyone who picked Antigravity for its multi-model choice. Free users lose the choice entirely, and paid Pro users lose models they are paying for, with no announcement beyond a footnote. The fix is concrete: switch Pro runbooks to the Claude 5.5 pair now, or move the Claude work to a harness where the model is still offered. See [Antigravity on AgentStackList](https://agentstacklist.com/agents/antigravity/) for the full setup.
Who it's for
Antigravity users on Free, Google AI Plus at $4.99, Google AI Pro at $19.99 and Ultra, and anyone who built a pilot on a free or trial seat.
How to try it
- Open the Antigravity model selector today and check what is actually listed. Google already moved this date once, from 2 November to 18 November, with no notice, so treat 18 November as the earliest credible date, not a commitment.
- On Free or Google AI Plus, plan around Gemini 3.8 Flash as the only model. It is a fast, capable default, but it becomes the only default, and the pricing page still advertises the tier as offering a variety of agent models.
- On Pro at $19.99, switch any runbook that names Claude Sonnet 4.6 or Opus 4.6 to Claude Sonnet 5.5 or Opus 5.5 before 18 November, and confirm your subscription is non-trial, because the 5.5 pair on Pro carries a footnote limiting it to non-trial subscriptions.
- If you need GPT-OSS-120b, note it was the only non-proprietary reasoning model in the selector. The open-weight path now means running it yourself through the Antigravity SDK with LiteRT, which needs more than 24GB of VRAM or unified memory, or hosting it elsewhere.
- Do not confuse this with Anthropic's own retirement calendar. Claude Sonnet 4.6 and Opus 4.6 are still served by Anthropic through Claude Code and the API; this is Google removing them from its own harness selector.
- Note: there is no changelog entry, no developer blog post and no deprecation-page row for this removal, so verify against the docs table for the tier you will deploy on, not the tier you tested on. Plan and availability: Antigravity model availability docs, footnote verified 2 and 3 November 2026.
Written by Emily Carter from the original post. Checked Checked 18 Nov 2026. Features, plans and availability can change, so confirm on the source before you rely on it.
antigravityclaudemodels
More Gemini tricks
- Antigravity CLI 1.5.14 streams live thinking and plans remote work before it edits · Nov 21
- Antigravity CLI 1.5.13 brings Gemini 4 Neon to ADC sign-in and stops surprise swarms · Nov 20
- Gemini CLI 0.69 adds a /stats heatmap for your daily tokens and custom themes · Nov 19
- Antigravity's /run-workflow turns one prompt into a Python script that runs parallel agents · Nov 17