Anthropic's next Sonnet model might be closer than we thought. According to leakers on X, partners are already testing a second checkpoint of Claude Sonnet 5.5, and a release could come as soon as this week.

Let's start with what's actually confirmed. When Anthropic launched Claude Opus 5.5 on September 22, it said: "Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks, with many of the same improvements to performance, efficiency, and safety." So the model is real. Anthropic just hasn't said when.

Claude Opus 5.5 launched on September 22. Image: Anthropic

Now the leaks. On Saturday, X user lyra, who has been posting Sonnet 5.5 details for the past week, claimed partners got a better build and that release is "upcoming week":

Partners got access to another Claude Sonnet 5.5 checkpoint, which is even better than the first one.
Release is upcoming week :* https://t.co/qhkJ7Kppmb

— lyra (@lyraxana) September 26, 2026

In an earlier post, the same account claimed Sonnet 5.5 has a 1M-token context window, 128K max output, adaptive thinking on by default, and pricing of $2 per million input tokens, $10 per million output tokens, and $0.20 per million for cache reads. It also claimed partners only got access for 24 hours.

Another leaker, Lumina, posted what it says is a Sonnet 5.5 listing from the Droid registry. It shows a slightly different context number:

🚨 Claude Sonnet 5.5 details from the Droid registry:

Don’t read into the credits as pricing btw:

• 872K context + 128K max output
• standard low -> max reasoning
• pretty much everything you would expect

It looks like a pretty serious upgrade too imo https://t.co/iHOimuxR0d pic.twitter.com/VQTXD45fdv

— Lumina (@LuminaBench) September 27, 2026

So that's 872K context in one leak versus 1M in the other. Either could be right, or the final release could differ from both. Lumina also warns not to read the credits shown as pricing.

Here's the part I find interesting. If the leaked $2/$10 pricing holds, Sonnet 5.5 would cost exactly the same as the current Claude Sonnet 5, which already has a 1M context window and 128K max output. In other words, the upgrade would be in how smart the model is, not in the price tag or the specs sheet.

That fits the pattern with Opus 5.5. Anthropic says it performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5 on typical workloads. A Sonnet that gets a similar jump at the same price would be a big deal for anyone running coding agents all day.

The timing matters too. OpenAI's DevDay is tomorrow, and dropping a new Sonnet the same week would be a very Anthropic move. As of Monday afternoon IST, though, there's nothing on Anthropic's news page or its X account, and Sonnet 5 is still the latest Sonnet in its docs.

Until Anthropic says otherwise, treat every number above as a rumour.