Originally published at devtoolpicks.com
Anthropic just announced three changes to Claude Code, all effective immediately.
First, Claude Code's five-hour rate limits are doubled for Pro, Max, Team, and seat-based Enterprise plans. Second, peak hours throttling is gone for Pro and Max accounts. Third, API rate limits for Opus models are raised substantially.
If you have been hitting rate limits mid-session or watching your agent stall during peak hours, this is the update you were waiting for.
What actually changed
Doubled five-hour rate limits. Every Pro ($20/month), Max ($100 or $200/month), Team, and Enterprise user now gets twice the number of messages per rolling five-hour window. Anthropic did not publish the exact new numbers, but if your previous limit was, say, 45 Opus messages per five hours, it is now 90. This applies to Claude Code specifically.
Peak hours throttling removed. Until today, Pro and Max users saw reduced rate limits during high-demand periods. If you have ever had Claude Code slow down or refuse messages at 2 PM Eastern on a weekday, that restriction is gone. Your rate limit is now the same at 3 AM and 3 PM.
Opus API rate limits raised. Developers building on the Claude API get higher rate limits for Opus models across all tiers. Anthropic published an updated rate limit table showing the new maximums per tier.
All three changes are live right now. No action needed. If you are on a paid plan, your limits are already updated.
Why this is happening now
Anthropic signed a deal with SpaceX to use the full compute capacity at Colossus 1, a data center in Memphis, Tennessee. That gives Anthropic access to more than 300 megawatts of new capacity and over 220,000 NVIDIA GPUs including H100, H200, and GB200 accelerators. The capacity will come online within the month.
This is on top of Anthropic's other compute agreements: up to 5 gigawatts with Amazon (nearly 1 GW by end of 2026), 5 gigawatts with Google and Broadcom (2027), $30 billion of Azure capacity with Microsoft and NVIDIA, and a $50 billion infrastructure investment with Fluidstack.
The SpaceX deal is notable for another reason. Elon Musk merged xAI into SpaceX earlier this year, and Musk has publicly criticized Anthropic on multiple occasions. The fact that SpaceX is now renting its flagship data center to a company Musk has called "doomed to become the opposite of its name" says a lot about how compute economics work in 2026. Revenue wins over rivalry.
What this means for Claude Code users
If you have been frustrated with Claude Code's rate limits, this is a meaningful improvement. Doubling the five-hour window means longer uninterrupted agent sessions. Removing peak hours throttling means you can work during normal business hours without your agent degrading.
Last month, Anthropic acknowledged that demand for Claude had led to "inevitable strain on our infrastructure" that was impacting reliability and performance during peak hours. If you read about the Opus 4.7 regression complaints, capacity constraints were widely suspected as a contributing factor. More compute directly addresses that.
For developers using Claude Code agents for longer tasks, this also reduces the risk of an agent burning through your rate limit mid-session and leaving your project in an incomplete state. A doubled limit means fewer interrupted workflows.
The peak hours change is arguably more important than the doubling itself. A high rate limit that gets cut in half during business hours is frustrating because that is exactly when most developers are working. Removing that restriction means your workflow is predictable. You know what your limit is regardless of when you are coding.
Who benefits most
Pro users ($20/month) get the biggest relative improvement. Pro was the plan most affected by rate limits and peak hours throttling. Doubling the cap and removing throttling makes Pro significantly more usable for serious development work. If you were considering upgrading to Max solely because Pro limits were too tight, it is worth re-evaluating.
Max users ($100 or $200/month) were already the highest-limit consumer tier. Doubling an already generous limit means Max users should rarely hit rate caps at all, except during extremely long autonomous agent sessions.
Team and Enterprise users get the same doubling. For organizations running multiple developers on Claude Code simultaneously, the combined capacity increase is substantial.
API developers building products on Opus models get higher rate limits across all tiers. If you have been batching API calls to stay under limits, check the updated table at platform.claude.com/docs/en/api/rate-limits to see if you can simplify your architecture.
How this stacks up against Codex
The timing is worth noting. OpenAI shipped the /goal command for Codex CLI last week, giving Codex a persistent workflow feature that competes directly with Claude Code's agent mode. Anthropic doubling rate limits the following week is not a coincidence. The AI coding tool competition is driving real improvements for users.
Codex operates on a usage-based model (you pay per task), so rate limits work differently. Claude Code's model charges a flat monthly fee with rate limits. Doubling those limits effectively halves the per-message cost for heavy users on Pro and Max plans. At $20/month for Pro with double the previous limit, Claude Code becomes harder to beat on value for developers who work within the rate cap.
The orbital compute footnote
Buried at the bottom of the announcement is a line about Anthropic expressing interest in "partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity." Data centers in space. Anthropic's reasoning is that compute required for next-generation AI systems is outpacing what terrestrial power, land, and cooling infrastructure can deliver.
This is speculative and years away from being real, but it signals how seriously AI companies are thinking about the compute bottleneck. The race is not just for better models anymore. It is for the physical infrastructure to run them.
FAQ
Do I need to do anything to get the new limits?
No. The changes are effective immediately for all paid plans. Your Claude Code rate limits are already doubled. No settings to change, no plan upgrade required.
What are the exact new rate limits?
Anthropic doubled the five-hour limits but did not publish the specific numbers for each plan tier. The API rate limit table for Opus models is available at platform.claude.com/docs/en/api/rate-limits.
Does this affect the free tier?
No. The doubling applies to Pro, Max, Team, and Enterprise plans only. Free tier limits are unchanged.
Will this fix the Opus 4.7 quality complaints?
Not directly. Rate limits and model quality are separate issues. But more compute capacity means Anthropic can run inference without the kind of resource constraints that may have contributed to inconsistent model behavior during peak hours. If some of the Opus 4.7 complaints were caused by infrastructure strain rather than model changes, more capacity should help.
Is the SpaceX partnership exclusive?
Anthropic gets all of Colossus 1's capacity, but this is one deal among many. Anthropic also runs on AWS Trainium, Google TPUs, and Azure. The company is hardware-agnostic and sources compute from multiple providers.
Bottom line
Doubled rate limits, no peak throttling, and higher API caps. These are the changes Claude Code users have been asking for since Anthropic admitted to infrastructure strain last month.
The SpaceX partnership is the infrastructure story behind the headline. 220,000 GPUs and 300 megawatts of capacity is a significant addition. Whether it is enough to keep up with demand as Claude Code adoption grows is a different question. But for today, if you are on a paid plan, your Claude Code experience just got noticeably better.
Top comments (0)