This week, Anthropic’s Cowork AI assistant shocked Wall Street. Now, Anthropic is refining the model and taking it even further.
Anthropic’s new Claude Opus 4.6 model announced Thursday is designed to make Cowork AI suitable for office work and coding tasks, potentially further raising concerns that AI tools could replace specialized software packages that companies use for those tasks.
Legal and financial analysis software stocks have fallen sharply in recent days, and the overall stock market has also fallen. The Nasdaq just posted its first two-day decline since April, dropping another 0.7% on Thursday.
Many experts wonder whether AI will eventually put some workers out of work. Tech giants like Anthropic, OpenAI, and Google are in a race to build the AI models that will power the workplace of the future.
Humanity rival OpenAI just introduced a new platform Thursday morning for creating AI agents that act like your colleagues. And Anthropic launched its Cowork tool in January.
It remains unclear whether investments in AI will benefit companies that adopt this technology. Anthropic is betting its new model will help it replicate the huge success of its Claude Code software outside of other types of office work.
“We think Opus 4.6 is a turning point in knowledge work in many ways,” Diane Penn, director of product management for research, told CNN in an interview ahead of the announcement.
One of those methods has to do with the way Opus 4.6 processes data. Anthropic says it has expanded Opus’ context window (the amount of information a model can remember at one time) from 200,000 tokens to 1 million tokens. A token is a unit of measurement that describes how an AI model understands text. The longer and more complex your query, the more tokens you will need.
By allowing Claude to process more information at once, Penn said it will be able to handle more complex tasks, such as making major changes to the entire codebase.
In Opus 4.6, Claude will be better able to know when to take more time to think through a request (a technique called reasoning) and when to respond immediately.
Anthropic also said the new model outperforms OpenAI’s GPT-5.2 model in benchmarks that assess how AI handles knowledge work in fields such as finance and law.
A new PowerPoint integration available in Research Preview allows users to create slides using Claude. AI can read layouts and fonts and create slides that match your desired corporate template. This was particularly difficult. Because unlike Excel, which is primarily data-driven, PowerPoint slides are Penn said this includes decisions about design elements such as color and text placement.
Anthropic says Claude Opus 4.6 handles documents, spreadsheets, slides, and other files that are closer to “production-ready” on the first try and require less human intervention.
Anthropic’s Opus 4.6 improves Claude’s non-technical tasks, but the update also includes improvements for software engineers. According to Anthropik, coding tasks can be divided into teams of agents rather than having one agent handle individual tasks, mimicking the way ergonomics teams operate.
The release comes after software stocks soared this week following the release last Friday of a plugin for Anthropic’s Cowork tool. These plug-ins, which allow users to customize Anthropic’s Cowork tools for specific industries such as law, finance, sales and marketing, raised concerns that the technology could replace specialized research and financial analysis software.
The exchange-traded fund (ETF) that follows the software industry had its worst day since April on Tuesday, falling nearly 6%. Thomson Reuters (TRI) fell 15.83% on Tuesday, while LegalZoom (LZ) also fell nearly 20% on the same day.
At the same time, there are growing questions and concerns about the role AI will play in the workplace, especially among entry-level technical roles. A report from Oxford Economics last year found that employment of new graduates in computer science and mathematics has fallen by 8% since 2022. A Google survey in September found that 90% of technology workers use AI in their jobs.
“These concerns are something we address and think about in every version of Claude and every product we offer,” Penn said, pointing to efforts such as Anthropic’s economic indicators report that studies the impact of AI on labor.
But eMarketer technology analyst Jacob Bourne said it’s still unlikely that these kinds of AI tools will reshape the job market. Security concerns can prevent many large companies from using these types of tools, as they often require permissions to access files and browsing activity. (Penn said Anthropic works with customers to ensure its technology meets security and IT requirements.)
“The panic about this is probably misplaced,” Vaughn said. “But I think this means traditional enterprise software providers need to continue to evolve.”
