Anthropic appears to be quietly developing two new advanced AI models, according to recently discovered web configuration files. The files contain references to “Claude Sonnet 4” and “Claude Opus 4,” which are likely part of the upcoming Claude 4 series and potentially represent the company’s most sophisticated AI systems to date.

The internal configuration files explicitly mention “Claude 4,” “Claude Sonnet 4,” and “Claude Opus 4,” suggesting a tiered approach to the new model lineup. These references include development notes such as “Not intended for production use” and “strict rate limits,” indicating the models are still in testing phases or limited early access.

Technical details in the files hint at advanced reasoning capabilities, with features like “show_raw_thinking” appearing in the configuration. This suggests the new models may offer enhanced logical processing compared to current versions.

Anthropic has scheduled a ‘Code with Claude’ event for May 22, though it remains unclear whether this event will include announcements about these new models or their capabilities.