Google Antigravity Cuts Quotas 92% While Claude Code Ships Opus 4.6 with 1M Context
Google slashes Antigravity free tier quotas by 92% and locks Pro users out for days. Meanwhile, Claude Code defaults Opus 4.6 to 1M tokens at the same price. A tale of two AI coding platforms in March 2026.
In the same week of March 2026, two AI coding platforms made moves in opposite directions.
Claude Code shipped Opus 4.6 with a 1M-token context window as the default — 5x more capacity at the same price.
Google Antigravity users reported multi-day account lockouts, a 92% free-tier quota cut, and pricing that pushes developers toward a $250/month plan.
The contrast tells a story about how these companies view their developer relationships.
What Happened at Google Antigravity
Google Antigravity launched in November 2025 as an agentic AI development platform. It promised generous rate limits and a free public preview. By March 2026, the developer community is in open revolt.
The Quota Cuts
Google made silent, unannounced reductions to Antigravity quotas:
| Change | Before | After | Reduction |
|---|---|---|---|
| Free tier daily requests | 250/day | 20/day | 92% |
| Gemini 2.5 Pro free tier | ~500/day | 100/day | 80% |
| Quota refresh cycle | 5 hours | 5-7 days | ~24x slower |
The most painful change: the quota refresh window. Developers expected their usage to reset every 5 hours. Instead, Google moved to a weekly calculation — meaning Pro users who hit their limit early in the week were locked out for days, not hours.
Pro Users Locked Out
On Google’s AI Developers Forum, the backlash was immediate:
“Before January I could use 300M input tokens/week; now I hit weekly limits at <9M input” — Google AI Developers Forum
AI Pro subscribers ($20/month) reported:
- 5-7 day lockouts after exceeding vaguely defined weekly limits
- No clear documentation on actual token quotas
- Terms like “high,” “generous,” and “meaningful” used instead of specific numbers
- The only way to unlock: buy AI credits ($25 for 2,500 credits) or upgrade to AI Ultra ($250/month)
The Pricing Wall
| Tier | Monthly | Daily Tasks | Concurrent | Reality |
|---|---|---|---|---|
| Free | $0 | 5 | 1 | Barely functional |
| AI Pro | $20 | 100 | 15 | Weekly lockouts reported |
| AI Ultra | $250 | ”Unlimited" | "Unlimited” | Only tier without caps |
The gap between AI Pro ($20) and AI Ultra ($250) is a 12.5x price jump with no middle option. For many developers, the message is clear: pay $250/month or deal with interruptions.
What Happened at Claude Code
On the same day Google developers were posting about lockouts, Anthropic rolled out Opus 4.6 with 1M tokens as the default context window for Claude Code.
The Upgrade
| Opus 4.5 | Opus 4.6 | Change | |
|---|---|---|---|
| Context window | 200K tokens | 1M tokens | 5x |
| Max output | 64K tokens | 128K tokens | 2x |
| Pricing | $5/$25 per MTok | $5/$25 per MTok | Same |
No beta headers required. No special flags. No extra charges for using the full 1M window.
Availability
- Max, Team, Enterprise plans: Opus 4.6 (1M) is the default model
- Pro plan: Opt in with
/extra-usage - API users: Just use
claude-opus-4-6— 1M context is automatic
Performance
On the MRCR v2 benchmark (needle-in-haystack retrieval at 1M tokens):
- Opus 4.6: 76% accuracy
- Sonnet 4.5: 18.5% accuracy
The model doesn’t just have more room — it actually uses it effectively. Context rot (performance degradation as context fills up) has been largely solved.
The Contrast
| Factor | Google Antigravity | Claude Code |
|---|---|---|
| Direction | Reducing quotas | Increasing capacity |
| Transparency | Vague limits (“generous”) | Clear specs (1M tokens) |
| Communication | Silent, unannounced cuts | Public release notes |
| Pro tier ($20/mo) | Weekly lockouts reported | Opus 1M with /extra-usage |
| Unlimited tier | $250/month | Max $200/month |
| Developer sentiment | Forum protests, The Register coverage | Generally positive |
The Pricing Math
To get uninterrupted AI coding:
- Google Antigravity: AI Ultra at $250/month (only tier without weekly caps)
- Claude Code: Max at $200/month (Opus 4.6 1M default, no caps reported)
For $50 less per month, Claude Code offers a clearly documented 1M context window with no surprise lockouts.
Why This Matters
The developer tools market is in a critical trust-building phase. Developers who commit to a platform are making a significant productivity bet — they’re building workflows, muscle memory, and team processes around these tools.
Google’s approach erodes trust:
- Silent quota reductions with no announcement
- Vague documentation that doesn’t match reality
- A pricing structure that feels like a bait-and-switch (free preview → aggressive monetization)
Anthropic’s approach builds trust:
- Clear, documented specifications
- Price stability (same $/MTok, more capacity)
- Transparent upgrade path (Pro → Max, clear differences)
The irony of a platform called “Antigravity” that keeps pulling developers down with quota walls hasn’t been lost on the community.
What Developers Are Saying
From Google’s AI Developers Forum:
- “Do they really think we wouldn’t notice a 92% free tier quota cut?”
- “Declared quotas don’t match actual limits. This is not transparency.”
- “Multi-day lockouts on a paid plan? That’s not a rate limit, that’s a service outage.”
From The Register (March 12, 2026): “Users protest as Google Antigravity price floats upward”
Meanwhile, the Claude Code community is mostly discussing what to do with 5x more context.
Our Take
Both platforms have different strengths. Antigravity offers model choice (Gemini, Claude, GPT-OSS) and an IDE-based workflow. Claude Code offers deep terminal integration and a focused Claude experience.
But on the question of how you treat developers who are paying you money — there’s no comparison this week.
One platform gave its users 5x more capacity at the same price. The other took away what it promised and asked for 12.5x more money to get it back.
Actions speak louder than marketing.
Sources: Google AI Developers Forum, The Register, Anthropic Models Documentation, Jules Usage Limits