The competition in AI development heats up again.

Last week, it seemed like the AI race was a two-horse contest between Anthropic and OpenAI, with both companies pulling ahead thanks to their powerful models, leaving competitors trailing. Google appeared to be lagging significantly, while Meta, despite pouring billions into revitalizing its AI efforts, had yet to make a mark as a serious player. Even the latest models from China, while impressive, seemed to be struggling to keep pace.

Then, in just a few days, the landscape shifted dramatically. Google and Meta both introduced cutting-edge models that they claim can compete directly with the best offerings from OpenAI and Anthropic. Suddenly, the competition is back on, and the outcome is anyone’s guess.

Until this week, Google seemed to have lost its footing among the AI elite. After launching Gemini 3 late last year, the tech giant had only managed to deliver minor updates. While Gemini 3 was a strong contender at its debut, it quickly fell behind newer models from OpenAI and Anthropic, as well as some emerging open-weight models from China.

Internal changes at Google further fueled the perception that the company was stepping back from the AI forefront. Veteran executive Jeff Dean left alongside several AI researchers to start a new venture. Demis Hassabis, the Nobel laureate who led Google’s AI powerhouse DeepMind, reduced his day-to-day involvement. Other AI teams were shifted from DeepMind to Google, indicating a significant shift in the company’s AI focus from London to California, leading to speculation that Google was prioritizing commercialization over research.

However, on Tuesday, Google turned that narrative on its head with the launch of Gemini 4 Argon, a new model that the company claims matches or surpasses flagship offerings from OpenAI and Anthropic across various benchmarks. Additionally, Google has set an aggressive pricing strategy for Argon, undercutting its competitors by as much as 80 percent during its introductory phase, and remaining significantly cheaper even after the promotion ends.

Meta’s comeback was equally striking. Unlike Google, which had long been the frontrunner in AI due to its resources and DeepMind, Meta had struggled to establish itself as a serious competitor. Despite its efforts, its models had failed to make a significant impact. Last year, the company initiated a major overhaul of its AI operations, investing billions in talent and infrastructure, including partnerships with Scale AI and its cofounder Alexandr Wang. In April, its new Superintelligence Labs launched its first major model, Muse Spark, which was impressive but not quite at the frontier.

That changed on Monday when Meta finally revealed a model that can compete with the industry’s best. According to the company’s tests, Muse Wonder Full can hold its own against leading models from OpenAI and Anthropic, including GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1, across various benchmarks. In essence, Meta has finally made its mark in the AI arena.

The resurgence of Google and the emergence of Meta have made the AI competition much more dynamic, highlighting how quickly the balance of power can shift in this space. Both companies have deep pockets and extensive ecosystems of products used by billions, advantages that newer players like OpenAI and Anthropic simply can’t match. The rapid success of Meta’s Muse agent assistant underscores this advantage, demonstrating the power of having billions of users at their fingertips. Meanwhile, OpenAI is working hard to build its own ecosystem, positioning itself as a viable alternative to Meta.

As the AI race enters a new phase, the push for more powerful models is becoming increasingly complex. Growing safety concerns and incidents involving rogue AI are prompting louder calls for a slowdown in development. Open models are also improving rapidly, making it challenging for any single company to maintain a lead for long.

If sheer power and performance are no longer sufficient to stay ahead, Google and Meta possess advantages that Anthropic and OpenAI lack. While they may have spent time catching up, recent developments show that it would be unwise to underestimate either of them.