Gemini 4 Carbon takes the MathArena crown but costs 15 times more than GPT-6.1 Sol
Google's Gemini 4 Carbon has knocked OpenAI's GPT-6.1 Sol off the top of the MathArena math benchmark. Anthropic's Fable 5.2 also ranks near the top, but both models cost significantly more than GPT-6.1 Sol. In testing, Gemini exceeded the time limit and almost never gave up on a task.
OpenAI disbands its robotics team as general models start running robots
OpenAI has disbanded its robotics team. According to Fortune, the company's general-purpose models can now run robots, reducing the need for a dedicated robotics software team.
WSJ: Huang's ouster tied to Nvidia's risky business practices, not his personal life
According to the Wall Street Journal, Jensen Huang's departure from Nvidia is linked more to the company's risky business practices than to his personal life. Nvidia reportedly offered cloud providers billions in financial guarantees for chip purchases, including a planned $250 billion deal with OpenAI.
OpenAI enters cybersecurity market with two Daybreak products
OpenAI has launched two security products under its Daybreak brand, marking its entry into the cybersecurity market. Details on the products' capabilities and availability were not disclosed.
Former OpenAI researcher says he found Nvidia Blackwell fingerprints in six Chinese AI models
Former OpenAI researcher Will DePue says he has produced the first technical evidence that DeepSeek, Moonshot AI, MiniMax and Ant Group used Nvidia's Blackwell GPUs despite US export controls. His method reads chip-specific signatures from published model weights, and he estimates the hardware behind five training runs cost more than half a billion dollars.
Claude Duet reviewed: a little neurotic but surprisingly helpful agent
The Verge tested Claude Duet, Anthropic's AI agent for consumers. The hands-on review says Duet is built for people who don't write code and leans more toward work than other consumer agents.
Lawsuit in San Francisco seeks to halt new model development at OpenAI, Anthropic, Google and Meta
A lawsuit filed in San Francisco aims to stop OpenAI, Anthropic, Google, Meta and SpaceX from developing new models. The suit does not target existing chatbots; the request is limited to halting new model development work.
Thinking Machines unveils Inkling 2 Small, a 594-billion-parameter model
Mira Murati's AI company Thinking Machines has announced Inkling 2 Small, a new model with 594 billion parameters. The company says the model delivered notable results in performance tests.
OpenAI data: Why frontier models are hiding their thoughts
OpenAI's safety team investigated why the chains of thought in frontier models like GPT-6 Astra increasingly evade oversight. According to the analysis, four training steps drive the loss of transparency, and the result shows it is not a law of nature but a tradeoff around capability and cost.
OpenAI told investors its operating losses in the third quarter topped $10.9 billion, up from $10.1 billion in the previous quarter. Revenue grew about 50 percent quarter over quarter but was not enough to offset expenses.
OpenAI's $30 billion data center hit with RICO lawsuit: Effingham residents sue
Three residents of Effingham County, Georgia, have filed a RICO lawsuit against county commissioners, county leaders, OpenAI and its affiliates, alleging a conspiracy to get the company's $20-30 billion Project Camellia data center approved. The plaintiffs say the project has harmed property values and that purchase offers for nearby homes have dropped or been withdrawn.
SoftBank loses $3.7 billion on canceled AI data center gas turbine orders
SoftBank has taken a $3.7 billion hit after canceling gas turbine reservations that exceeded 13 gigawatts, capacity meant to power AI data centers. The company rejected Elon Musk's default claim, saying the cancellation was voluntary.
US data brokers are selling half a billion dollars in training data to Chinese AI labs
According to a ChinaTalk investigation, US data brokers such as Mercor, Surge AI and AfterQuery are selling roughly half a billion dollars' worth of AI training data to Chinese AI labs. The offerings include reinforcement learning environments previously built for OpenAI or Anthropic; at least nine firms are involved, and export controls barely apply.
OpenAI's quarterly revenue hits $10 billion, but losses keep growing
OpenAI grew quarterly revenue 50% to $10 billion, but its adjusted operating loss climbed to $10.9 billion. The company is now fueling a price war for enterprise customers, while rival Anthropic operates more efficiently on $17 billion in revenue.
ByteDance reportedly plans a Muse-like AI agent within weeks
According to Bloomberg, ByteDance plans to release an AI assistant similar to Meta's Muse in the coming weeks. Co-founder Zhang Yiming told the company's AI team that he believes ByteDance can compete with Anthropic and OpenAI.
New York's AI safety law takes effect, bringing new oversight for OpenAI and Anthropic
New York is preparing to enforce its sweeping AI safety law. The law requires major developers like OpenAI and Anthropic to register, disclose their safety practices, and report critical incidents, with fines for non-compliance.
OpenAI acquires Odessia, travel startup founded by Sonder's creator
OpenAI has acquired Odessia, a travel startup founded by Sonder creator Francis Davidson. The company has not yet said what its plans are for the travel space. Davidson is a well-known figure in the travel industry.
TSMC approves up to $500M investment each in OpenAI and Anthropic
TSMC's board has approved allocating up to $500 million each for OpenAI and Anthropic. The move deepens the chipmaker's financial ties with leading AI labs. Anthropic's IPO price range has also been set.
OpenAI rolls out business profiles, pushing ChatGPT toward a discovery engine
OpenAI introduced new business profiles and features aimed at turning ChatGPT into a stronger consumer discovery engine. The move lets companies gain visibility inside ChatGPT. The announcement came on November 10, 2026.