{"text":[[{"start":7.9,"text":"Microsoft is releasing a series of new AI models in a bid to catch up with leading AI lab Anthropic, whose focus on business users poses a threat to the company’s dominant software products."}],[{"start":19.15,"text":"The company’s AI chief Mustafa Suleyman told the FT that Microsoft’s superintelligence team was “less concerned” with the consumer-oriented approach taken by leading labs at Google, Meta and OpenAI. "}],[{"start":32.25,"text":"“We’re more focused on the Anthropic-style which is enterprise [use cases], developers and coding,” he said. “That’s the journey we’ve been on.” "}],[{"start":40.45,"text":"The former Google DeepMind co-founder was speaking ahead of Microsoft’s Build developer conference on Tuesday, at which he unveiled seven AI models, including one focused on reasoning that the company claims is comparable in coding abilities to Anthropic’s Opus 4.6 model released in February."}],[{"start":61.6,"text":"Suleyman said Microsoft was “now at the absolute frontier” before conceding that Anthropic, which has released two higher-performing models since Opus 4.6, was still ahead by several months."}],[{"start":74.5,"text":"“We’ve closed an enormous gap in six months.”"}],[{"start":77.65,"text":"Microsoft has been pursuing “true self-sufficiency” in AI after it signed a deal in late October that restructured its relationship with OpenAI, allowing the start-up to complete a corporate restructuring. Microsoft retains a 27 per cent stake in the ChatGPT maker and has kept access to its most advanced models until 2032. "}],[{"start":98.95,"text":"The company previously relied on its access to OpenAI’s advanced models to power its full suite of consumer and enterprise-oriented AI products. "}],[{"start":108.15,"text":"Satya Nadella, Microsoft’s chief executive, has made a series of bets aimed at supporting the group’s shift away from OpenAI, including committing to invest up to $5bn in Anthropic in November as part of a $30bn cloud computing deal."}],[{"start":123.85000000000001,"text":"Suleyman, who joined Microsoft in 2024 from his start-up Inflection as part of a $650mn deal, has seen his role overhauled in the past year as he focuses squarely on the group’s bid to build state-of-the-art models. "}],[{"start":138.15,"text":"Yet Microsoft continues to lag leading AI labs and has struggled to convince the market that it is generating sufficient return on vast capital expenditure while facing pressure in the software market. "}],[{"start":149.5,"text":"Anthropic’s Cowork, an AI coding tool for business users that requires no coding abilities, triggered a market sell-off at the beginning of the year after it unveiled a series of add-ons targeting white-collar work. The release led to fears over the future of enterprise software groups, with Microsoft’s shares trading down 10 per cent year-to-date."}],[{"start":169.65,"text":"Microsoft’s AI lab on Tuesday also unveiled an “ultra efficient” coding model that it said was fine-tuned for the group’s GitHub developer platform."}],[{"start":179.25,"text":"Suleyman said that the combination of a coding and reasoning model — which can break down complex tasks into smaller pieces — would help the software giant build “thinking and coding” agents, autonomous bots that can carry out tasks for users, that would represent a significant boost for business customers."}],[{"start":199,"text":"He added Microsoft’s model development would also help drive down its costs over time, as the group sacrifices “significant margin” to Anthropic when serving its products to customers. “It translates into real dollars on the bottom line.”"}],[{"start":219.25,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1780454035_6339.mp3"}