The competition in the generative AI market has entered a new phase.
Elon Musk and Mark Zuckerberg, who until recently ranked behind companies like OpenAI and Anthropic, are now narrowing the gap with the leaders by relying on heavy investments, extensive infrastructure, and new models.
SpaceX this week released the Grok 4.6 model; a model that, in the Intelligence Index of Artificial Analysis, is roughly on par with OpenAI's GPT-5.6 Sol Max and evaluated only slightly below Anthropic's Fable 5 Max.
This performance is particularly important for the AI project led by Musk because it shows Grok has been able to noticeably reduce its distance from the market’s advanced models.
Musk has also promised that the next version of this model, Grok 4.7, will arrive within three to four weeks.
He has claimed that this version, after further training with a massive volume of SpaceX data, could surpass all current models. That claim, however, remains a promise and prediction for now, and it will be necessary to wait for the release of the new model and independent test results to evaluate it.
On the other side, Meta is pursuing a different strategy. Instead of focusing solely on achieving the highest benchmark scores, the company is trying to offer models with competitive performance at much lower cost.
One example of this approach is Muse Glimmer; an open-weight model designed so small that it can even run on a laptop.
Mark Zuckerberg also explained this approach in a wide-ranging statement about the future of AI with the slogan “superintelligence for everyone,” and he again emphasized Meta’s ambition to develop AI with broad accessibility.
This change of course comes after Meta made significant adjustments to its AI structure and strategy following the disappointing performance of Llama 4.
In line with this, Zuckerberg has invested $14.3 billion in Scale AI and has brought that company’s CEO, Alexandr Wang, into the fold to lead Meta’s new efforts.
Musk has taken an even more aggressive approach, merging xAI into SpaceX and agreeing to acquire Cursor for $60 billion, bringing computational power, SpaceX’s proprietary data, and a growing AI-powered coding platform together in one group.
The importance of these developments is not limited to the raw power of the models. The cost of using AI has become a decisive factor in the competition.
Companies and users do not necessarily need the most powerful and expensive model on the market, and models that can deliver performance close to leading models at lower cost can capture a significant share of the market.
Despite these advances, it cannot yet be said that Musk and Zuckerberg have fully caught up with OpenAI and Anthropic. Analysts still place those two companies in the leading position, and some of their more powerful models have not yet been released publicly.
According to AXIOS, one venture investor also emphasized that it cannot be said these two companies have completely caught up with OpenAI and Anthropic. A Meta employee told the media that the company still feels behind leading labs in model performance and development.
Meanwhile, Google’s situation could also change the competitive equations. The company, previously known as one of the strongest tech giants to challenge OpenAI and Anthropic, has faced increased pressure following the departure of several prominent AI figures, including its chief scientist, and some issues around the Gemini model.
Such circumstances could create more opportunities for SpaceX and Meta to solidify their positions in the public AI market.
Overall, AI competition is shifting from a purely technical race to a multifaceted contest that simultaneously involves model strength, cost, computational infrastructure, access to data, and financial capacity.
OpenAI and Anthropic remain leaders, but the capital and economic scale of SpaceX and Meta have allowed Musk and Zuckerberg to once again draw nearer to the competition.