Antigravity Adds Claude Opus 5.5 and Sonnet 5.5 for Pro and Ultra, and Retires GPT-OSS and Claude 4.6 on November 2

On October 3, 2026, an Antigravity user posted that Claude Opus 5.5 and Sonnet 5.5 were arriving on the Pro and Ultra plans, that GPT-OSS 120B was being removed, and that the change might hint at an imminent Gemini 4 Argon rollout. The tweet drew replies about trial accounts, usage limits and confusion over who qualifies.

Most of the factual core can be checked against Google's own documentation, which makes this a story with a confirmed skeleton and a rumor-filled margin. This post separates the two.

TL;DR — what people are asking

What Google's own page says

Antigravity's models documentation lists the available reasoning models by provider. As read on October 3, the relevant entries are:

Three points stand out.

1. The 5.5 models are gated by plan. They are listed for Pro and Ultra only, and for Pro only on non-trial subscriptions.
2. Free and Google AI Plus users lose Claude access on November 2, on the page as written. The only Claude models listed for those plans are the 4.6 models scheduled for removal. This is an inference from the table, not a statement Google made, so confirm it against your account.
3. Gemini 4 Argon is not listed. The documentation does not mention it.

Pricing context: pricing pages for Antigravity describe plans in terms of rate limits and credits rather than per-token prices. For reference, Epoch lists Opus 5.5's API price at $4 per million input tokens and $20 per million output tokens, but that is Anthropic's API pricing, not what Antigravity charges you; inside Antigravity you spend plan quota.

The "my Pro plan says no" problem

The tweet's author asked why an interface said the models would not be available on a plan that was already Pro. The documentation offers a plausible answer: the Pro listing says non-trial subscriptions only. Replies in the thread guessed the same, asking whether the account was a promo account and observing that students and trial plans appear to be excluded.

A Google AI Developers Forum thread posted the same morning describes a Pro user who did not see Opus 5.5 and received a message that third-party models will be removed. The only reply was another user advising patience, and I found no Google staff response in the thread. So the situation is:

• Likely: trial, promotional or student-style Pro plans are treated differently from paid, non-trial Pro.
• Possible: a staged rollout, so some paid accounts see the models later than others.
• Unknown: Google's exact wording of the in-app warning and whether it applies to your account.

If you pay full price for Pro and still see a warning, sign out and back in, update the app, and contact Google support with a screenshot. Do not assume the models are permanently unavailable.

What is rumor

"This hints that Gemini 4 Argon is close"

It is speculation. What explainx.ai's own coverage found is that Argon is rolling out first to trusted defenders through a program called Fairwind, with broader access promised to paid API customers and Google AI Ultra subscribers and no calendar date; see Gemini 4 Argon: launch, benchmarks and pricing. Swapping which third-party models appear in Antigravity says nothing about when Google's own flagship ships. A model-lineup cleanup and a first-party launch are separate events.

"The five-hour limit now equals the weekly limit"

One user complained that the five-hour window now seems equal to the weekly limit and hoped it was a bug. Another reported that a single prompt on Pro consumed 16 percent of their quota. Both are anecdotes from the Antigravity 2.0 app, and neither is documented.

What is documented, via third-party pricing summaries from September, is that paid plans refresh quota every five hours up to a weekly ceiling, that the free tier resets weekly, that unused allowance does not roll over, and that real limits depend on available capacity. Google does not publish absolute numbers at any tier. One summary of the pricing page even flagged conflicting signals about which plan gets third-party models, and advised checking the model selector in a signed-in account. A heavy reasoning model such as Opus 5.5 plausibly drains a quota faster than a Flash model, so higher per-prompt consumption would not be surprising, but that is an expectation, not a measurement.

Why the model swap is a big deal

Claude Opus 5.5 is currently the top model on Epoch's Capabilities Index at 167, narrowly ahead of GPT-6 Astra; our post on Opus 5.5 topping the Epoch index covers what that number does and does not mean. Getting it inside a Google-hosted coding agent changes the trade-offs between harnesses:

• Access without a separate Anthropic subscription. If you already pay for Google AI Pro or Ultra, Opus 5.5 becomes a model you can pick inside the same app, subject to quota.
• A real comparison. The same model in different harnesses can behave differently. Compare Opus 5.5 in Antigravity with Opus 5.5 in Claude Code on one of your tasks, looking at turns taken and tokens or quota consumed.
• Choosing between 5.5 siblings. Sonnet 5.5 is the cheaper and faster option; our Opus 5.5 versus Sonnet 5.5 comparison helps decide which to use for which work.
• Less reason to use GPT-OSS. The removal of GPT-OSS-120b ends a free open-weights option inside the product; if you used it for cheap bulk edits, plan a replacement such as a Flash model.

What to do before November 2

1. Open the model selector in your signed-in account and see what is listed. Documentation and rollout can differ.
2. Confirm your plan type. Check whether your Pro plan is a paid, non-trial subscription or a promo, trial or student arrangement.
3. Audit your defaults. Find any workflow, config or saved agent that names Claude Opus 4.6, Sonnet 4.6 or GPT-OSS-120b, and change it before the removal date.
4. Test the 5.5 models on your tasks. Use a handful of real tickets and compare quality and quota use against your current default.
5. Watch quota for a week. Note how much a typical prompt consumes at each model, so you know your real headroom.
6. If you are on Free or Plus, decide now whether the Gemini models cover your needs, since Claude access appears to end on November 2.
7. Keep a fallback. If your work depends on a particular model, make sure you can run it somewhere else, such as Claude Code.

For background on the product itself, our guides cover the Antigravity CLI features, the migration from Gemini CLI to Antigravity, and Antigravity remote control. For how practitioners compare harnesses, see the agent harness tier list reaction.

What people are asking

Is this an official announcement?

The source tweet is a user's reading of in-app notices. The models documentation, which I fetched, corroborates the main points. I did not find a Google blog post announcing the change.

Will the new models be free?

No. The page lists them for paid Pro and Ultra only.

Can I keep using Claude 4.6 after November 2?

Not in Antigravity, according to the removal dates on the models page.

Does a Workspace or Business plan get these models?

The page I read does not mention Claude availability for enterprise plans, so check with your administrator.

Is Antigravity 2.0 the same as the IDE?

One user in the thread wasn't sure whether their experience was Antigravity 2.0 or the IDE. The models page doesn't distinguish between them, so check the model selector in whichever one you use.

Honest limitations

• I relied on Google's models page and pricing summaries as fetched on October 3, 2026; documentation can change or lag the product.
• The tweet and forum posts are user reports, and I could not verify their screenshots.
• Usage-limit statements are secondhand and may be out of date.
• The inference that Free and Plus users lose Claude access on November 2 comes from the table and has not been confirmed by Google.

Bottom line

Antigravity's documentation lists Claude Opus 5.5 and Sonnet 5.5 for non-trial Google AI Pro and for Ultra, and schedules Claude 4.6 and GPT-OSS-120b for removal on November 2. The Argon hint and the quota complaints are not confirmed. Check your selector, your plan type and your defaults now, and test the 5.5 models on real work before you rearrange a workflow around them.

Related on explainx.ai

• Claude Opus 5.5 tops the Epoch Capabilities Index
• Claude Opus 5.5 launch: benchmarks and pricing
• Opus 5.5 vs Sonnet 5.5
• Opus 5.5 task cost in Claude Code
• Gemini 4 Argon: launch, benchmarks and pricing
• Antigravity CLI features
• Gemini CLI to Antigravity migration guide
• Agent harness tier list reaction

Sources: Antigravity models documentation · Antigravity pricing · Google AI Developers Forum thread, October 3, 2026

Model availability, plan rules and removal dates reflect Google's documentation as read on October 3, 2026 and may change.