Anthropic launches Claude Opus 4.7 as its most intelligent public AI model

Anthropic continues its rapid product releases in 2026, launching Claude Opus 4.7 this Thursday.

As the latest in a series of hybrid reasoning models, Opus 4.7 is now the most advanced AI system made publicly available by Anthropic. It offers notable improvements in multi-step reasoning and coding capabilities.

However, the company stated in a press release that the model is intentionally less powerful than Claude Mythos, an unreleased system considered too risky for public use.

The new model is immediately accessible via the Claude AI platform, the official API, and enterprise partners such as Microsoft Foundry. While the pricing remains the same as the previous version, users should be aware of potential increases in token usage.

Since the system performs deeper, more effortful thinking, it consumes more output tokens. To assist developers, Anthropic has released a detailed migration guide with strategies to optimize token use and manage processing costs.

According to the company’s official blog post, the updated system offers substantial enhancements in visual intelligence, document analysis, and, in particular, complex coding tasks.

Anthropic claims that artificial intelligence is now more refined and creative in handling professional workloads, consistently producing higher-quality interfaces, presentations, and documents.

Early users report handing off extremely difficult programming assignments to the model with remarkable confidence, noting that it handles long-running tasks with rigorous consistency. The system pays precise attention to user instructions and independently devises internal mechanisms to verify its own outputs.

A detailed model card reveals that while the system lags behind the restricted Claude Mythos, it competes fiercely with frontier models from OpenAI and Google.

On Humanity’s Last Exam, the system outperformed all public competitors without external tools, although it lagged slightly behind GPT-5-4-Pro when those tools were introduced.

Importantly, Anthropic highlights that this model has a low risk profile concerning misaligned behaviors.

The company has reported significant reductions in severe hallucinations and critical informational omissions, demonstrating that it is more reliably honest than its immediate predecessors in the rapidly evolving artificial intelligence sector.