OpenAI's new GPT-5 model — the engine that will power its key products, including the free and paid versions of ChatGPT — throws down a gauntlet to its arch-competitor Anthropic.

The big picture: Anthropic's Claude is frequently viewed as the best AI for writing code, which could be the reason OpenAI leaned so heavily on GPT-5's coding skills as the company introduced its new model yesterday.

Catch-up quick: GPT-5 is OpenAI's first model to combine a traditional large language model with reasoning capabilities that allow AI engines to achieve better results by using more computing time to address a task.

• The company also promised GPT-5 will hallucinate less and be better at health-related tasks, among other improvements.
• GPT-5 will be less sycophantic than its predecessors, according to OpenAI.
• Pietro Schirano, CEO of AI startup MagicPath, called GPT-5 the best coding model out there and a great collaborator. "This is electricity arriving in every home," he posted on X, "a before and after moment for how we build."

Yes, but: The new model arrived months later than originally expected and critics say it doesn't bring the world much closer to the kind of AI that can solve the climate crisis or cure cancer.

• "It will not harness the incredible power of artificial intelligence as a tool to solve humanity's most pressing challenges," Future of Life Institute president Max Tegmark said in a statement. "Instead, it will pour fuel on the fires set by its predecessors: destroying jobs, upending economies, and turbocharging scams and deepfakes."

The other side: For casual users of ChatGPT, the upgrade is likely to feel significant.

• "The people discussing this release on Twitter will be disappointed in a first reaction, but 99% of people using ChatGPT are going to be so happy about the upgrade," Nathan Lambert wrote in his Substack newsletter.
• The potential new crop of users could create a fresh wave of experimentation and novel use cases, which is what happened when OpenAI increased access to ChatGPT's image generation capabilities.

Between the lines: OpenAI spent much of an hourlong livestream yesterday highlighting GPT-5's coding abilities, showing demos, touting a range of benchmarks and providing comments from a range of early testers, including Cursor, Vercel and Windsurf.

• "Our team has found GPT-5 to be remarkably intelligent, easy to steer, and even to have a personality we haven't seen in any other model," Cursor CEO Michael Truell said in a statement. "It not only catches tricky, deeply-hidden bugs but can also run long, multi-turn background agents to see complex tasks through to the finish."
• CEO Sam Altman told reporters that GPT-5 allows for on-the-fly custom software. "This idea of software on demand will be a defining part of the new GPT-5 era," he said.

By the numbers: Coding prowess is driving massive revenue growth at Anthropic.

• Anthropic is generating revenue at a nearly $5 billion-per-year pace, according to The Information. That's up from a $4 billion-per-year pace earlier this month, and reflects its status as the go-to choice for programmers and coding apps.
• OpenAI, meanwhile, is currently at a $12 billion per year revenue run rate, a figure that reflects the company's broader business and larger size.

Zoom out: While OpenAI heavily emphasized GPT-5's coding capabilities, the new model is replacing a host of older models, including those that power the free and paid versions of ChatGPT.

• Users will also be able to customize GPT-5's personality as part of a feature in research preview. The four initial options are "Cynic," "Robot," "Listener" and "Nerd."
• Altman insisted that once you use GPT-5 it is hard to go back to an earlier model, likening it to reverting to a lower-resolution display after you've grown used to an iPhone's Retina display.

What we're watching: GPT-5 isn't AGI, but improved coding could woo more developers, while fresh personalization options and reduced hallucinations could attract more everyday users to the free version of ChatGPT.

The Federal Trade Commission should investigate Meta's investment in Scale AI, a dozen consumer advocacy and competition groups wrote to the agency yesterday.

Why it matters: The Trump administration wants to be both tough when it comes to antitrust enforcement and hands-off when it comes to AI development.

What they're saying: "This transaction follows Meta's familiar anticompetitive buy or bury strategy," the groups write in a letter to FTC chair Andrew Ferguson shared exclusively with Axios.

• "The structure of the transaction — a non-controlling stake that falls just below the 50% threshold for mandatory reporting — is a deliberate attempt to evade FTC scrutiny."
• "This is a familiar playbook which includes key aspects typical of Big Tech's minority investments in generative AI companies."
• Signers of the letter include Public Citizen, the Tech Oversight Project, the Consumer Federation of America and the American Economic Liberties Project.

Context: Meta announced in June it was investing $14.3 billion in Scale AI, acquiring a 49% stake in the company and also hiring a number of top staffers.

• A Meta spokesperson said at the time of the announcement that "as part of this, we will deepen the work we do together producing data for AI models, and [former Scale CEO] Alexandr Wang will join Meta to work on our superintelligence efforts."
• Sumit Sharma, executive director of NextGen Competition: "The harms extend well beyond model development and aim squarely at cementing Meta's dominance in adjacent markets such as social media and surveillance advertising."
• "Meta already faces an antitrust lawsuit in these markets. Do we really want Mark Zuckerberg and Meta to have even more control over our information ecosystem?"

The groups say the deal is a de facto vertical acquisition meant to exploit gray areas in the law.

• They note that in a January report on AI partnerships the FTC "warned that such deals could give dominant firms access to sensitive technical and business information of their partners."

• President Trump called on Intel CEO Lip-Bu Tan to resign, following concerns about his reported ties to Chinese companies. (Axios)
• Apple has lost about a dozen people from a key AI team, with 50 or 60 remaining on the team that builds foundation models, per the Financial Times.
• "I am a failure. I am a disgrace." Google's Gemini bot has been spouting self-doubt to users as the result of what one Google exec called "an annoying infinite looping bug." (Business Insider)
• I missed this when it was announced earlier this week, but former FCC chair Jessica Rosenworcel is taking over as executive director of MIT's Media Lab, starting next month.

Peacock dropped the first trailer for "The Paper," the newspaper-themed series from the creators of "The Office" that debuts in September.

• This tweet from Axios colleague Andrew Solender summed up my expectations: "this is going to be the worst representation of journalism ever committed to screen and i will be watching every minute of it."