Antigravity adds Gemini 3.5 Flash Low with ~45% fewer tokens

• Antigravity added Gemini 3.5 Flash (Low) across the IDE, CLI, and desktop app, and ai_for_success's feature summary says it uses about 45% fewer tokens than Flash Medium.
• The same feature summary claims Flash Low generally beats Gemini 3 Flash High on SWE tasks, which makes this look like a cost-control tier, not just a fallback model.
• Google staff spent the weekend resetting quotas repeatedly: OfficialLoganK's update said weekly limits were reset, kevinhou22's follow-up announced another reset a day later, and Kevin Hou's later post said quotas were reset across free and paid plans.
• The quota churn followed visible user frustration, with ai_for_success's complaint describing a five-hour lockout after light image and avatar use, while koltregaskes' post complained about constant permission prompts in Antigravity 2.0.

You can see the new low tier in ai_for_success's rollout post, the repeated quota resets in OfficialLoganK's update and Kevin Hou's later note, and the other quiet weekend changes in Kevin Hou's changelog-style post, which also mentions a 2x context bump for Gemini 3.5 Flash and major Windows bug fixes.

The clearest product change is a new Gemini 3.5 Flash (Low) option inside Antigravity. According to ai_for_success's rollout post, it is available in the IDE, CLI, and desktop app.

That post lists three concrete claims:

• about 45% fewer tokens than Gemini 3.5 Flash Medium
• availability across IDE, CLI, and desktop
• better SWE-task performance than Gemini 3 Flash High

Google acknowledged the token complaint directly. In the repost captured by ivanleomk's repost, @_mohansolo says the team heard concerns that Antigravity was consuming too many tokens for simple tasks, then added Flash Low in response.

Quota management changed almost as fast as the model menu. Kevin Hou's post says quotas were reset across all plans, paid and free, while OfficialLoganK's earlier update said the team had already reset everyone's weekly limits, and Kevin Hou's day-later update announced yet another reset.

The user reports explain why this became the story. In ai_for_success's complaint, a heavy Gemini user says one image edit and a single avatar prompt triggered a five-hour usage limit, with video generation failing in the same session.

There is also a separate, more favorable limit change in the background. rohanpaul_ai's summary says every paid Antigravity tier now gets permanent 3x Gemini rate limits, which sits awkwardly next to the same week's reset notices and complaints about new usage-based caps.

Antigravity 2.0 also appears to have shipped with enough interaction friction that users started asking where the promised automation went. In koltregaskes' post, a user says all permissions were set to Allow but the app still surfaced "1,000" approval prompts, and also complains that picking an option still requires a separate Submit click.

At the same time, Google increased the breathing room for longer runs. Kevin Hou's changelog-style post says Antigravity doubled max context length for Gemini 3.5 Flash, and andyzhang's repost echoes the same point while tying it to users hitting compaction repeatedly.

Part of the confusion was basic product labeling. OfficialLoganK's update says the IDE in Antigravity 2.0 was never removed, and that the UI was changed to make the distinction between AGY 2.0 and AGY IDE clearer in the top-right corner.

That same cleanup pass included two more concrete fixes from Kevin Hou's weekend update:

• clearer in-app separation between AGY 2.0 and AGY IDE
• major bug fixes, especially on Windows
• another quota reset on top of the earlier one