Just last week, Anthropic and OpenAI appeared to have secured an unassailable lead in the AI competition, advancing with ever-more-capable models that left competitors trailing. Google looked irretrievably behind, while Meta — despite pouring billions into reviving its struggling AI aspirations — had failed to prove itself a genuine threat.
Even China’s latest impressive offerings seemed merely to be chasing the leaders.
Now Google and Meta are mounting a direct challenge to Anthropic and OpenAI at the cutting edge of AI. In just a few days, the landscape shifted dramatically. Both companies debuted frontier models they assert can match the top offerings from OpenAI and Anthropic head-to-head. The contest has reignited overnight — and the outcome is wide open.
Until this week, Google appeared to have permanently ceded its position among AI’s elite. Following Gemini 3’s release late last year, the tech behemoth offered only marginal upgrades. While Gemini 3 was formidable at launch, its advantage proved short-lived, and it faltered against subsequent releases from OpenAI and Anthropic — even some open-weight models from China.
A wave of internal upheaval only reinforced the impression that Google was withdrawing from the AI vanguard. Longtime executive Jeff Dean left alongside several AI researchers to found a startup.
Demis Hassabis, the Nobel laureate who led Google’s AI powerhouse DeepMind for years, relinquished daily operational control. Additional AI teams were transferred from DeepMind into Google’s core organization.
Collectively, these moves indicated a fundamental relocation of Google’s AI center of gravity from London to California, sparking conjecture that the company was shifting focus from research to commercialization.
Then, on Tuesday, Google shattered that narrative by unveiling Gemini 4 Argon, a new frontier model the company says equals or surpasses flagship models from OpenAI and Anthropic across multiple benchmarks — whatever significance those benchmarks still hold.