Anthropic on Tuesday rolled out Claude Sonnet 4.6, claiming the model delivers “human-level” performance on some computing tasks.
In a blog post, the startup said early users saw improvements in everyday office work, including navigating spreadsheets, completing forms and consolidating data across multiple browser tabs.
However, it admitted that the model can’t yet compete with the “most skilled” humans at using computers.
Anthropic said the new version of Sonnet is stronger at coding, handling longer documents and completing multi-step workflows. In developer testing, users preferred the new model over its predecessor, saying it delivered more consistent results and followed instructions more accurately.
The company’s safety evaluations found Sonnet 4.6 to be “as safe as, or safer than” its other recent models, with improved resistance to prompt injection attacks, where malicious actors attempt to hijack the model by hiding instructions on websites.
Anthropic rolled out its “frontier” Opus 4.6 model earlier this month. The company said Sonnet 4.6 approaches the performance of Opus, with benchmark results showing improvements over its predecessor.
Sonnet is now the default model across Claude’s free and paid tiers, with pricing unchanged. The free plan has also been upgraded to include file creation, the ability to connect to other apps, and longer conversations before hitting usage limits.
