Anthropic raises Claude Code usage limits, credits new deal with SpaceX

Last month, Anthropic reportedly signed massive deals with Microsoft, Google, Amazon, Nvidia, and more to scale up its access to compute infrastructure. The company credits those, along with this SpaceX deal, for its ability to raise limits, though gains from some of them will take time to materialize.

Is the increase of credits a gimmick? I don’t know. With some companies you think you are receiving a reasonable number of credits but then the credits vanish before you can finish whatever you are doing. Of course, not all. Some are reasonable.

The announcement was accompanied by an increase in those usage limits; Anthropic doubled Claude Code’s five-hour window limits for Pro and Max subscribers, removed the peak-hours limit reduction on Claude Code for those same accounts, and raised API limits for its Opus model. The table below outlining the Opus changes was shared in the company’s blog post on the topic.
Key Features of the Free Tier:

Models: Access to capable models such as Claude 3.5 Sonnet and Haiku.
Artifacts: A side window for viewing rendered code, websites, and SVGs.
Usage Limits: Limits are based on 5-hour sessions and can be reached quickly during peak times (5 am–11 am PT).
Data Usage: Free conversations may be used for model training.
Capabilities: Includes file uploads (PDF, Word, CSV) up to 30 MB for analysis, as well as image recognition.

Free Access Alternatives:

Claude Code: Can be used entirely free via OpenRouter, which allows connecting to various free models.
Mobile Apps: Free access is available on iOS, Android, and web.

Differences from Pro:

The Pro plan offers about 5x more usage than the free tier.
Pro users get access to the most capable model (Opus 4.6), which is not available to free users.
Pro/Team users can opt-out of data training.

Note: Free tier users may experience faster depletion of limits during high-traffic time.

So, like I said the number of credits isn’t really impressive if you burn through those credits the moment you fart or the systems farts. For example on Chatgpt I have burned through credits when I couldn’t get an image right or when trying to deal with censorship issues.