Sources for every figure, price, quote and product claim this video puts on screen. Checked 2026-08-21.
The plan is listed publicly as Individual, $0/month. The same plan card lists Unlimited Tab completions, Unlimited Command requests, and Basic weekly rate limits.
Agent model access on that free plan is listed as Gemini 3.5 Flash, Gemini 3.1 Pro, Gemini 3 Flash, Claude Sonnet & Opus 4.6, gpt-oss-120b. That is the source for the claim that the model list holds Gemini, Claude and GPT OSS, and for the claim that Claude is in the free tier rather than only in a paid one.
Source: https://antigravity.google/pricing
The plans documentation describes the free tier as “Meaningful quota, refreshed weekly” with a “Weekly rate limit”, against Google AI Pro’s “High, generous quota, refreshed every five hours until weekly limit reached” and Google AI Ultra’s “The highest, most generous quota, refreshed every five hours” plus “Access to third-party models”.
The capacity line the video quotes is verbatim from the same page:
The baseline rate limits are primarily determined to the degree we have capacity, and exist to prevent abuse.
The page also states that usage limits are subject to modification to manage system capacity, and that additional credits can be purchased by Pro and Ultra subscribers but not on the free plan. That is the basis for the video saying the higher plans get more generous limits and credit pools while the free plan does not.
Source: https://antigravity.google/docs/plans
Antigravity is described as having two surfaces rather than a chat sidebar: an Editor View, with tab completions, inline commands and an agent in the side panel, and an Agent Manager, an agent-first surface for spawning and overseeing several agents across workspaces at once.
The artifact types named in the launch post are task lists, implementation plans, walkthroughs, screenshots and browser recordings. The video names task lists, implementation plans, screenshots and browser recordings, all four of which appear on that list.
A browser recording is documented as a step-by-step recording of the agent working through a UI, which the developer watches to verify the result rather than running the app themselves. That is the basis for the video’s claim about moving the review surface from logs to artifacts.
Sources:
The three figures the video reads out as the prices developers have been trained to accept are Anthropic’s own consumer tiers:
| Plan | Price | Claude Code |
|---|---|---|
| Free | $0 | not included |
| Pro | $20/month billed monthly ($17/month billed annually) | included |
| Max 5x | $100/month | included |
| Max 20x | $200/month | included |
Claude Code is listed as a Pro feature and above, and is not part of the free Claude tier. That is what makes the comparison in this video a real one: the terminal product has no free tier of its own.
Source: https://claude.com/pricing
The video is explicit that Anthropic’s Claude Code did not itself become free, and that what changed is that Claude models are reachable inside a free Google workspace. Nothing here supports a stronger claim than that, and the script does not make one.