Google and Meta push back into AI frontier race
After Anthropic and OpenAI seemed to hold the lead, Google and Meta introduced frontier models that they claim can match the top competitors. Google had been criticized for only incremental progress since Gemini 3, and Meta had yet to prove itself despite large AI investments. The new releases have made the contest at the AI frontier competitive again.
Google’s momentum had faded after Gemini 3, with later updates seen as modest. Internally, Jeff Dean and other researchers left to start a company, Demis Hassabis reduced his daily role, and some AI groups were reassigned from DeepMind to Google. That fueled talk of a center of gravity moving from London to California and toward commercialization.
Meta had spent large sums on people and infrastructure, including Scale AI and cofounder Alexandr Wang. Its Superintelligence Labs released Muse Spark in April, which drew attention but did not reach the top tier. Meta then launched a frontier model on Monday; Google followed Tuesday with Gemini 4 Argon.
The renewed competition could benefit developers, businesses, and consumers by expanding access to capable models and possibly lowering prices, as Google’s introductory discount suggests. It may also pressure Anthropic, OpenAI, and others to release stronger systems faster. Workers and the public might encounter more AI features in everyday products, while questions about reliability, safety, and market concentration may grow alongside adoption.