Grok 4.6 and Meta's Muse Glimmer Narrow the Gap With OpenAI and Anthropic

Elon Musk's xAI and Mark Zuckerberg's Meta released new AI models this week that narrow — though don't close — the performance gap with OpenAI and Anthropic, according to Axios reporting citing third-party benchmarks and industry analysts.

The numbers

xAI's Grok 4.6 scored essentially even with OpenAI's GPT-5.6 Sol Max on the Artificial Analysis Intelligence Index, a closely watched third-party benchmark aggregator, and just behind Anthropic's Fable 5 Max, according to Artificial Analysis data cited by Axios. Specific score values were not disclosed in available reporting.

Musk said on X that Grok 4.7 will ship in three to four weeks and claimed — his word, not an independently verified result — that it will "exceed all current models" after additional training on SpaceX proprietary data.

Meta's release, Muse Glimmer, is an open-weight model small enough to run locally on a laptop. Meta has not disclosed pricing, parameter count, or full benchmark scores for the model in materials reviewed for this article. The company is positioning it as evidence that competitive performance no longer requires frontier-scale cost.

Why it matters

Both companies spent heavily to get here. Zuckerberg overhauled Meta's AI division after the underwhelming Llama 4 launch, invested $14.3 billion in Scale AI, and hired Scale's CEO Alexandr Wang to run the effort. Musk folded xAI into SpaceX and agreed to acquire Cursor for $60 billion, combining SpaceX's compute and proprietary data with a fast-growing AI coding platform.

Zuckerberg framed Meta's push as "superintelligence for everyone" in a manifesto published this week, reviving the company's open-source AI strategy amid renewed competition with Chinese labs.

The skepticism

Not everyone is convinced the gap has meaningfully closed. "I wouldn't say they've caught up to OpenAI or Anthropic," said John Belton, a portfolio manager at Gabelli Funds. A Meta employee told Axios the company still feels behind frontier labs on model performance and development timelines. OpenAI and Anthropic are also reportedly holding back systems deemed too sensitive for public release, including work related to cybersecurity risks that delayed at least one OpenAI model.

Sonali Basak, chief investment strategist at iCapital, argued that price — not just raw benchmark performance — is becoming a decisive factor for many users. "You don't need to be the highest performing, most expensive model in the market to be able to be successful," she told Axios. She added that the more consequential question is whether Musk's and Zuckerberg's AI investments can "fuel the core businesses" that financed the comeback.

The wider context

The timing coincides with turbulence at Google, which had positioned itself as the tech giant best equipped to challenge OpenAI and Anthropic. An exodus of senior AI talent, including Google's chief scientist, along with setbacks tied to its Gemini model line, has created what Axios describes as a potential opening for Meta or xAI to claim status as Wall Street's public-market AI leader.

What this means

No independently verified benchmark score, pricing detail, or parameter count for Grok 4.6, Grok 4.7, or Muse Glimmer has been published outside company statements and the Artificial Analysis index cited here. The claims of near-parity with GPT-5.6 and Fable 5 Max should be read as directional signal, not confirmed technical parity — frontier labs still control the top end of the performance curve and are reportedly withholding more capable systems. The real story is strategic: Musk and Zuckerberg are using balance-sheet scale (SpaceX's infrastructure, Meta's ad revenue) to compress cost curves rather than chase raw benchmark leadership, a bet that shifts competition toward distribution and price rather than pure capability.