Anthropic’s next big AI model may be coming very soon. Leaked code from Anthropic‘s internal systems has revealed references to a new model called Opus 4.7. Developers spotted the mentions in commit messages, which are notes programmers leave when updating code. Reports from The Information also confirm that Opus 4.7 is being prepared as the company’s next flagship model. Some insiders say it could drop as early as this week.
The leaked code includes a code name linked to Opus 4.7 development: capybara. Internal API references also suggest the model was spotted close to its release date. That’s a strong signal that the launch is close.
Leaked code reveals the codename capybara, signaling Opus 4.7’s release may be just days away.
Opus 4.7 is expected to bring several improvements over the current top model, Opus 4.6. Users can expect better prompt accuracy, stronger consistency, and more realistic details in outputs. The new model may also fix rate limit problems that have frustrated some Claude users. Opus 4.6 currently leads Anthropic’s lineup with a 1 million token context window and hybrid reasoning that allows both quick and extended thinking.
Alongside Opus 4.7, Anthropic is also reportedly preparing a new AI-powered design tool. This tool would let users build websites, presentations, and landing pages just by describing what they want. It’s seen as a direct threat to companies like Gamma, which already offers AI-based design features. The tool may also challenge Google’s own AI design efforts.
The design tool is expected to launch around the same time as Opus 4.7. Anthropic is also said to be rolling out a full redesign of its Claude desktop app alongside these releases.
Anthropic’s current models are available through its Claude API, as well as Amazon Web Services Bedrock and Google Vertex AI. The model lineup includes Opus 4.6, Sonnet 4.6, and Haiku 4.5. Each handles text, images, and multiple languages. Opus 4.6 remains the strongest option for complex tasks like coding and enterprise work. Recent benchmark tests show Opus 4.6 dropped from 2nd to 10th place on the hallucination benchmark, falling from 83.3% to 68.3% accuracy.
The pace of Anthropic’s releases shows the company isn’t slowing down anytime soon. Comparisons to Google Stitch suggest the new design tool is squarely aimed at non-technical users who want to create polished content without writing a single line of code. The broader AI industry is also experiencing rapid expansion, with the healthcare AI market alone projected to grow from $11 billion to $187 billion by 2030.
References
- https://www.youtube.com/watch?v=kTbHKcX-5es
- https://www.youtube.com/watch?v=MXZ9cfkYuRI
- https://www.anthropic.com/claude/opus
- https://platform.claude.com/docs/en/about-claude/models/overview
- https://www.anthropic.com/transparency
- https://www.theinformation.com/briefings/exclusive-anthropic-preps-opus-4-7-model-ai-design-tool
