Anthropic introduces AI that can work independently for up to 7 hours - Claude Opus 4
At the first Code with Claude conference, Anthropic announced a new flagship model - Claude Opus 4, which it calls "the best AI model for programming in the world".
Here's What We Know
Opus 4 is able to perform complex multi-step tasks independently for up to seven hours in a row, which is a significant step in the development of agent systems - AI capable of acting without human intervention.
Compared to previous models, Claude Opus 4 works with multiple tools simultaneously, follows instructions more accurately, and is 65% less likely to resort to tricks or workarounds during tasks. Combined with improved memory and local file access, the model is ready to become the basis for the autonomous digital agents of the future.
Along with Opus 4, the company also introduced an updated Sonnet 4 model that combines fast response with deep thinking. Sonnet 4 received the same improvements as Opus and is available even to free users.
These models are already integrated into Visual Studio Code, JetBrains, and GitHub Copilot, and the Sonnet 4 API remains priced at $3 for one million incoming tokens and $15 for outgoing tokens.
This week has been crucial for the AI industry: Google introduced AI Mode for all US search users, and OpenAI announced the acquisition of hardware startup Jony Ive for $6.5 billion. As part of the deal, the io team of about 55 specialists, including former Apple employees, will join OpenAI to develop new hardware devices with integrated artificial intelligence.
Source: Anthropic
