Anthropic launches Claude Sonnet 4.6 with computer use upgrades

Anthropic launched Claude Sonnet 4.6, which is an upgrade that improves coding skills, computer use, agent planning and knowledge work.

In a nutshell, Anthropic said you can throw Claude Sonnet 4.6 a full project. Anthropic said that early access users prefer Sonnet 4.6 to Claude Opus 4.5. 

Claude Sonnet 4.6 is the default model in Claude.ai and Claude Cowork and available on Anthropic's free and Pro plans. 

Anthropic’s LLM releases have lead to a giant SaaS freakout on Wall Street. The working theory is that Anthropic will upend every category of enterprise software. 

• How three hotel giants are playing agentic commerce, AI
• How CEOs are answering the dreaded LLM disruption question
• SaaS under the microscope: How to evaluate your vendor
• SaaS death knell storyline illogical, but margin compression is here
• Observability stands out in enterprise software wreckage

Box CEO Aaron Levie said Sonnet 4.6 is an upgrade that delivers a 15% jump in performance and accuracy in Box's evaluation. 

On X, Levie said:

"We’ve been testing the model with Box AI on a variety of complex knowledge work tasks across many different sectors with enterprise content. These tasks approximate real work scenarios where a user needs to read through many documents to complete tasks like generating reports, doing due diligence, providing expert analysis, and more. 

In the public sector, we saw a jump from 77% to 88% in accuracy for complex tasks; healthcare saw a jump from 60% to 78%; and legal saw 57% to 69%. We saw similar levels of jumps across a range of other industries as well."

According to Anthropic, Claude Sonnet 4.6 excels in computer use and has graduated from experimental. "There are no special APIs or purpose-built connectors; the model sees the computer and interacts with it in much the same way a person would: clicking a (virtual) mouse and typing on a (virtual) keyboard," said Anthropic, which noted that Sonnet 4.6 trails human-level computer use capabilities but is improving. 

On benchmarks, Claude Sonnet 4.6 performs better than flagship models on agentic financial analysis and office tasks and is competitive in other areas. 

Separately, Figma said it is bringing work in Claude Code into Figma.

• Anthropic's Claude Code revenue doubled since Jan. 1
• How CEOs are answering the dreaded LLM disruption question
• Anthropic launches Claude Opus 4.6, OpenAI rolls out GPT-5.3-Codex
• Anthropic tries to head off NIMBY with electric bill missive
• Anthropic: Claude AI will be ad free
• Anthropic paper on disempowerment in actual AI usage