Investors are lowering their expectations for Anthropic’s upcoming stock market debut, with some now doubting it can reach a $1.5 trillion valuation. The shift follows a tech stock selloff and concerns about heavy AI spending.
An independent assessment by METR found that Anthropic’s AI models were improving on measurable tasks at roughly twice their earlier pace as of October 2026. The researchers said AI-assisted research likely drove much of the increase, but their estimate remains uncertain and Anthropic withheld some supporting evidence.
OpenAI has reportedly secured $11 billion in an unfinished funding round that values the company at $1.25 trillion. The round could reach about $20 billion when it closes in the coming weeks, according to the report.
Perplexity has introduced teammates, AI agents that handle ongoing work through its Computer assistant. They can coordinate with each other and work in the background, but are currently available only to Max subscribers, who pay $200 a month.
Thinking Machines has launched Inkling 2 Small, which it says ranks first among Western models slated to make their weights available. It scores much higher than its predecessor on AI tests but costs more per task, and its weights have not yet been released.
Crusoe confirmed it sold an unfinished Wyoming data center project to Google, which will take over construction. Google plans to invest $17 billion in the site; the sale price was not disclosed.
Amazon is letting select U.S. customers text Alexa to find, compare and buy products, with wider access planned in the coming weeks. The launch follows Amazon’s decision in September to block Meta’s Muse from searching its site.
A reported ChatGPT connector issue sent a confirmation code from one app through another without telling the user. The server rejected the request because the code belonged to the first app, but the report warns that a similar switch could expose data to another provider.
Abdulrahman Alfozan pleaded not guilty to a charge that he placed a recording device in a meeting room at OpenAI’s San Francisco office. He remains in jail without bail and could face up to five years in prison if convicted.
Anthropic examined about 207,000 research sessions to see how scientists use Claude. Researchers used it most for data analysis and writing, with heavier use in computer science and math than in lab-based fields.