Google's Gemini 4 Argon, Meta's Muse Wonder Full reset AI race

The shift caps a period in which Anthropic and OpenAI looked like they had the AI race to themselves. Google's last major release, Gemini 3, launched late last year and was formidable at the time, but its lead was fleeting and it struggled against later models from OpenAI, Anthropic, and even open-weight models out of China. A series of internal shake-ups reinforced the impression Google was stepping back: Jeff Dean left with other AI researchers to launch a startup, Demis Hassabis stepped back from day-to-day leadership of DeepMind, and other AI teams were moved out of DeepMind into Google proper, signaling a center of gravity shifting from London toward California.

Meta's path was different. It had never really established itself at the frontier despite trying, and last year it began an expensive overhaul of its AI operations, spending billions on talent and infrastructure, including Scale AI and its cofounder Alexandr Wang. Its Superintelligence Labs released Muse Spark in April — impressive, but not quite frontier. Muse Wonder Full, unveiled Monday, is the model Meta says finally reaches that level. Google's Tuesday release of Gemini 4 Argon, with aggressive per-token pricing, answered the same question in a single week.

The two companies' return changes the shape of the competition. Both have deep pockets and vast ecosystems of products used by billions of people, advantages OpenAI and Anthropic cannot match, as Meta's Muse agent assistant shows. At the same time, the drive to develop ever more powerful models is becoming increasingly fraught, with safety concerns and rogue AI incidents fueling louder calls to slow development, and open models improving. That suggests raw performance alone may no longer be enough to hold a lead for long.