OpenAI’s new reasoning models: What you need to know

Keep the emails coming; I love to hear from you. (Simply reply to the newsletter.)

On my end, I was reminded of a passage by Hunter S. Thompson—yup, I’m going there—that all too easily fits some of tech’s leading lights:

“There he goes. One of God’s own prototypes. A high-powered mutant of some kind never even considered for mass production. Too weird to live, and too rare to die.”

Alright, enough of that for now. Today’s news below. —Andrew Nusca

Want to send thoughts or suggestions to Fortune Tech? Drop a line here.

OpenAI debuts new ‘reasoning’ models and coding agent

OpenAI has released two AI “reasoning” models that it says are its most capable yet as the company seeks to gain a lead over its rivals. 

It also released an open-source AI agent that helps computer programmers code. Called Codex CLI, it marks the first time since 2019 that OpenAI has introduced a significant open-source tool.

The new AI models are the full-scale version of its o3 model, which OpenAI says is its most advanced AI system, as well as a smaller, but more efficient model called o4-mini. 

OpenAI said o3 and o4-mini have the ability to reason directly about visual information, such as sketches, diagrams, or photos—even ones that might be blurry or of poor quality.

The models will be immediately available to users of its paid ChatGPT Plus and Pro services, as well as organizations that use its enterprise-focused Teams and API products.

The release of the new models comes at a time when OpenAI faces pressure to show it remains at the forefront of AI development. 

Earlier this year, China’s DeepSeek upended conventional wisdom about the technological edge U.S. AI labs such as OpenAI enjoyed for years. DeepSeek’s open-source R1 mimicked the “chain of thought” reasoning that OpenAI’s proprietary o-series models offer. 

OpenAI also faces increased competition from other proprietary model providers. 

In February, Anthropic became the first to offer a model that combines quick, intuition-like answers with the ability to also perform “chain of thought” step-by-step reasoning if a prompt requires it. And last month, Google unveiled its Gemini 2.5 Pro model, a reasoning model that beat OpenAI’s o3-mini model on numerous benchmarks. —Jeremy Kahn

TikTok was ‘highly urgent’ threat to Meta, Zuckerberg says

On the third day of the FTC v. Meta antitrust trial, Meta CEO Mark Zuckerberg testified that TikTok posed a “highly urgent” competitive threat to his company when it arrived on the scene in 2018.

“We observed that our growth slowed down dramatically” as TikTok gained traction, Zuckerberg said Wednesday. “It was highly urgent. This has been a top priority for the company for several years.”

On Monday and Tuesday, Zuckerberg spent several hours answering questions from the attorney for the FTC.

The U.S. agency seeks to demonstrate that Meta monopolized the social media category by buying up the competition, including Instagram and WhatsApp, in the previous decade and creating “network effects” that discouraged users from leaving the company’s ecosystem.

On Wednesday, Zuckerberg answered questions from his company’s own attorney. 

The chief executive said Meta competes with an array of different platforms, not all of them conventional social media: Alphabet’s YouTube, Apple’s iMessage, Microsoft’s LinkedIn, Elon Musk’s X, and so on. 

Meta’s position is that there’s no monopoly with such a broad competitive landscape. The FTC’s position is that Meta competes narrowly with Snap’s Snapchat when it comes to sharing information with friends and family.

On Wednesday, the presiding judge asked Zuckerberg if network effects still matter today. 

“These apps now serve primarily as discovery engines,” the CEO replied. “People can take that content to messaging engines.” —AN

Spotify suffers a three-hour widespread outage

Spotify is back online after experiencing a major outage affecting thousands if not millions of users across North America on Wednesday.

According to Downdetector, which lets users report which services are not working, reports of Spotify being down spiked shortly before 9 a.m. ET. Reports piled up over the next two hours.

Spotify did not mention the root cause of the issue, but pointed to its X account that was posting hourly status updates about the situation. 

The company said “all clear—thanks for your patience,” shortly after noon ET.

In December 2024, Spotify reported having 675 million monthly active users around the globe, which included 263 million subscribers who pay for Spotify Premium. 

2024 was also the first year Spotify achieved its first full year of profitability, with net profits around $1.17 billion, compared to around $549 million in losses in 2023. —Dave Smith

More tech

—Tech stocks drop 7%. Nvidia, AMD warn of higher costs from trade war; ASML misses on order expectations given uncertainty.

—Perplexity x Samsung. The AI company is reportedly in talks to integrate its assistant with the Korean giant’s devices. 

—Chris Krebs to resign from SentinelOne. Trump’s former cybersecurity chief digs in after his ex-boss directs the Justice Dept. to investigate him.

—What’s going on at BluSmart? Service suspension at the Indian EV ride-hailing startup as regulators investigate Gensol, which shares cofounders.

—SoftBank, OpenAI mull UK Stargate. If the $500 billion U.S. project proves successful, why not extend it across the pond?

—AI music is slowly taking over. One in five songs uploaded to Deezer are AI-generated, company says.

—Temu, Shein cut ad spending on U.S. social media. Down 31% on Meta, down 19% on X.

—Lyft buys German taxi app FreeNow. The service was acquired from BMW and Mercedes-Benz for $197 million.

Endstop triggered