Anthropic claims Claude Sonnet 4.6 showcases ‘human-level capability’ in multi-step tasks.
Anthropic has said that developers prefer its latest Claude Sonnet 4.6 to its predecessor, the Sonnet 4.5, “by a wide margin”. A majority of the users, it claimed, liked the new model even over Opus 4.5, the company’s latest frontier model.
The model launch comes just after Anthropic announced a $30bn Series G raise earlier this month led by Coatue Management and Singapore’s GIC. The round took the AI giant to a post-money valuation of $380bn – more than doubling its value from the last round it announced in September.
AI models are leaping bounds as their creators push out newer advances at increasing speeds. However, the pace of these advancements has accelerated a massive sell-off in SaaS stocks in recent months. AInvest reports that the collapse in software stocks is a “full-blown sector-wide rout”.
iShares Expanded Tech-Software Sector ETF is down by about 21pc year-to-date, while major companies, including ServiceNow, Salesforce, Adobe, all had their shares dragged down in recent weeks as fears of AI disruption in the sector takes over.
Claude Sonnet 4.6 isn’t quelling those fears, with Anthropic boasting that the new model shows a “major improvement” in computer use skills, compared to prior Sonnet models. The company first introduced computer use with Claude 3.5 Sonnet and Claude 3.5 Haiku back in 2024.
The new model, Anthropic said, showcases “human-level capability” in tasks such as navigating a complex spreadsheet or filling out a multi-step web form.
According to early users, Sonnet 4.6 reads context more effectively, is less prone to overengineering and “laziness”, and is “meaningfully better” at instruction taking. These users have also reported fewer false claims of success, fewer hallucinations and more consistent follow-through on multi-step tasks.
Overall, the new model approaches Opus-level intelligence at a lesser price point, Anthropic said. Sonnet 4.6 is comparable to Opus 4.6 in agentic coding, agentic computer use and agentic tool use, while being better at agentic financial analysis and office tasks.
The model is available on all Claude plans, including the free tier, which is now by default Sonnet 4.6. According to Anthropic, evaluations suggest that Sonnet 4.6 is “overall” safe, and safer than its recent Claude models.
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.