Anthropic is secretly working on new models called Claude Sonnet 4 and Opus 4, which are believed to be the company’s most advanced AI models.
This is according to Anthropic web configuration files, which were recently updated with references to Claude Sonnet 4 and Opus 4.
As you can see in the above image showing the internal configuration file, Anthropic has specifically mentioned “Claude 4,” “Claude Sonnet 4,” and “Claude Opus 4.”
These appear to be part of the Claude 4 series, which we think could be Anthropic’s smartest models yet.
The references suggest Claude Sonnet 4 and Opus 4 are being tested or promoted internally, with some warnings like “Not intended for production use” and “strict rate limits,” meaning they’re still under development or early access.
There are also references to features like “show_raw_thinking,” which point to reasoning capabilities.
On May 22, Anthropic is hosting a ‘Code with Claude’ event, but we don’t know if the event is related to the new models.
Based on an analysis of 14M malicious actions, discover the top 10 MITRE ATT&CK techniques behind 93% of attacks and how to defend against them.
