← Back to Home
AI NEWS 2026-11-16

AI news, 16 November: OpenAI model helps tackle decades-old physics puzzle

OpenAI’s GPT-6 Astra aids candidate Standard Model lattice proof. Physicist Nathaniel Craig of UC Santa Barbara, working with OpenAI, described how GPT-6 Astra took the lead in developing the detailed mathematical proof for a candidate nonperturbative lattice definition of the Standard Model—the core theory of particle physics. Craig supplied the initial strategy for representing force fields on successively finer spacetime grids; the model helped construct the proof and perform Lean formal checks. The work, featured as the launch post of OpenAI’s new science blog, addresses a problem open for more than 25 years involving chiral fermions on a lattice. Experts reviewed it and it remains a candidate definition requiring further physical consistency tests, not a fully settled result. The development highlights AI’s growing role in fundamental scientific research beyond coding or writing assistance.

Anthropic’s IPO draws heavy demand ahead of Nasdaq debut. Orders for Anthropic’s offering of 825 million shares at $70–$78 each reached roughly six times oversubscription, according to reports citing sources familiar with the process. Pricing is expected around 18 November with trading under the ticker ANT on 19 November, potentially valuing the Claude maker near $1.4 trillion fully diluted. A concurrent $26 billion private placement includes major stakes from Nvidia, Google and others. The company has reported strong recent revenue growth, though executives have noted uncertainty around future compute costs. The deal ranks among the largest tech IPOs ever and tests investor appetite for frontier AI labs.

Alibaba launches Qwen4-Max with lower costs and safety upgrades. Alibaba released its new flagship model Qwen4-Max to general API access, featuring a mixture-of-experts design, native vision, a 1-million-token context window, and list prices cut about 25% to $1.50 input / $4.50 output per million tokens. The company says internal evaluations show average agent task costs falling over 60% versus the prior generation, with stronger performance on professional and scientific work. Safeguards against misuse and certain code vulnerabilities were tightened. Smaller Qwen4 models are slated for open weights later. The launch intensifies competition on price and capability from Chinese labs.

US and allies sign declaration to secure AI supply chains. The United States and eight partners—including Japan, Australia, the UK, South Korea, India, Israel, Qatar and the UAE—signed the San Francisco Declaration under the Pax Silica initiative, agreeing to jointly identify and protect critical AI technologies and supply chains (semiconductors, minerals, energy) viewed as vulnerable to China, the Financial Times reported. Germany, Sweden, the Netherlands and the European Commission are expected to join soon. Working groups and investment hubs are planned. The move comes amid ongoing export-control debates and ahead of high-level US-China talks.

Nvidia leadership uncertainty continues. Reports indicate former CEO Jensen Huang has discussed a possible return with allies days after resigning amid personal and security-related controversies, adding volatility ahead of earnings. A public petition seeking his reinstatement while investigations proceed has gathered over 100,000 signatures, including some employees and investors. Interim leadership under CFO Colette Kress is in place; Asian partners have sought and received supply reassurances.