If you've been hitting rate limits on Claude Code, you've probably noticed something changed in May 2026: the limits roughly doubled. You can run more coding sessions, handle larger codebases, and hit fewer "please wait" messages.
The reason is unexpected: Anthropic leased Colossus 1 — the Memphis supercomputer built by xAI (now SpaceXAI) with 220,000+ Nvidia GPUs. More compute capacity means higher rate limits for users.
Key Takeaway
Claude Code rate limits approximately doubled in May 2026. If you previously hit limits during long coding sessions, try again — the experience has improved significantly. Pro ($20/mo) and Max ($100/mo) tiers both benefit.
What Are the New Limits?
Anthropic hasn't published exact rate limit numbers (they adjust dynamically based on load), but community reports confirm roughly 2x the previous capacity:
| Tier | Before May 2026 | After May 2026 |
|---|---|---|
| Claude Pro ($20/mo) | Moderate — hit limits in ~2-3 hour sessions | Roughly doubled — longer sessions before limits |
| Claude Max ($100/mo) | Higher — still hit limits on heavy days | Roughly doubled — heavy usage sustainable |
| API | Tier-based rate limits | Higher throughput for paid tiers |
How to Maximize Your Claude Code Usage
Use CLAUDE.md files. Corrections saved in CLAUDE.md persist across sessions, so you don't waste tokens re-explaining project context. One good CLAUDE.md can save 30-50% of your token budget per session.
Use Claude Projects for persistent context. Upload reference files (style guides, architecture docs, API specs) so they load automatically. This is more efficient than pasting context into every session.
Prompt efficiently. Vague prompts waste tokens on wrong approaches. Specific prompts land on the right solution faster. The Prompt Optimizer restructures any coding prompt for precision — fewer iterations means fewer tokens means more coding within your limits.
Use the right effort level. Set effort to "xhigh" for complex tasks and "high" for standard coding. Don't use "max" for everything — it consumes more tokens with diminishing returns on simpler tasks. See our Claude 4.7 prompting guide for details.
---📬 Want more like this? We track Claude updates that affect your workflow. Subscribe free →
---Frequently Asked Questions
Why did Anthropic lease from SpaceX?
SpaceXAI (formerly xAI) has excess compute capacity — using only 11% of Colossus for Grok. Rather than build more data centers, Anthropic leased existing capacity at favorable terms. Faster and cheaper than building from scratch.
Will limits keep increasing?
Likely. Anthropic's 80x revenue growth funds infrastructure expansion, and the SpaceX compute deal is just one source. As capacity grows, rate limits should continue improving — especially if Anthropic wants to maintain subscriber growth.
Is Claude Max ($100/mo) worth it now?
If you code with Claude Code daily and previously hit limits regularly, yes. The doubled limits at the Max tier make heavy daily usage sustainable. If you only code occasionally, Pro at $20/mo is still sufficient with the new limits.
Disclosure: Some links in this article are affiliate links. We only recommend tools we've personally tested and use regularly. See our full disclosure policy.