Claude’s Quality is Dropping – Here’s Why

Claude's performance has taken a noticeable hit lately. I've been using it daily for several months, and the degradation is pretty clear. Let me break down what's happening and why I think it's occurring.

Current Issues

If you're using Claude regularly, you've probably seen these problems:

• Constant rate limiting and error messages
• Being forced into "concise mode" due to "high capacity"
• Getting significantly worse responses than before
• Responses getting cut off mid-generation
• Severely degraded performance with long inputs

That last point is particularly concerning. Claude was known for handling long documents better than any other AI - it was one of its standout features. Now it seems to ignore large chunks of input text or fails to properly synthesize information from longer documents. What used to be Claude's specialty has become notably worse.

The Official Line

Anthropic's CEO Dario Amodei recently addressed this on Lex Fridman's podcast. He insisted they never switch to lower-quality models during high load. But what's interesting is what he didn't address - there are many other ways to reduce compute load without technically switching models.

The Real Deal

While they're not swapping in a worse model, they're likely using various optimization techniques to handle the load:

• Reduced sampling parameters
• Lower precision computation
• Quantization
• Shallower beam search
• Possibly truncating or compressing long inputs

This matches what we're seeing - same model, worse outputs. It's particularly noticeable with long inputs, where the model seems to struggle maintaining context and attention across the full text. It's a way to handle more users without technically lying about not switching models.

Why Now?

The timing makes sense. Microsoft just integrated Claude into Copilot, which probably brought a massive influx of users. Claude was already known as one of the best models available, and now it's getting hammered with traffic from multiple sources.

I've noticed the quality drop is especially bad during peak hours. Sometimes you'll get a great response, but often it feels like you're getting a half-baked version of what Claude used to provide. Try giving it a long document now - the difference from a few weeks ago is stark.

What's Next

Amazon's $4B investment in Anthropic should help with scaling, but that'll take time to implement.

The real question is how long this will last. Scaling AI infrastructure isn't easy, and even with billions in funding, it'll take time to build out the necessary compute capacity. Until then, we're likely stuck with this degraded version of what used to be the best AI assistant available.

What I'm building

Delegate tasks. Get software.

Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.

Take a look at vroni.com