Anthropic has released a new generation of artificial intelligence models, Claude Opus 4 and Sonnet 4, which are focused on coding, complex logic tasks, and long-term autonomous operation.
The company calls the Claude Opus 4 its most powerful model. It can work continuously on a task for several hours – during testing, the model worked autonomously for seven hours straight. According to Anthropic, the Opus 4 outperforms the Google Gemini 2.5 Pro, OpenAI GPT-4.1 and o3 reasoning models in coding tasks and working with tools such as web search.
Claude Sonnet 4 is a general-purpose model that replaces Sonnet 3.7. It features improved accuracy, reasoning, and code handling while remaining more accessible than Opus.
New features include "thinking summaries," which summarize chains of reasoning for easier viewing, and an "extended thinking" mode that lets you switch between thinking and using a variety of tools to get more accurate answers.
Opus 4 and Sonnet 4 are already available through Anthropic’s API, Amazon Bedrock, and Google Cloud Vertex AI. Paid users have access to both models and extended thinking, and Anthropic has also said it plans to update the models more frequently to stay competitive with OpenAI, Google, and Meta.
Recall that Anthropic also recently introduced a new API that will allow Claude to perform real-time web searches and provide up-to-date information for programs and agents built on these models.