Elon Musk Says Grok 4.7 Lands in 10 Days and Will Beat Every Model

Elon Musk said Grok 4.7 will be released to the public in 10 days and that the model will surpass every AI model currently available.

The release follows Grok 4.6, which SpaceXAI shipped on August 12. It also comes as OpenAI announced that its new Astra model will be launched soon.

Follow us on X to get the latest news as it happens

Grok 4.7 comes out in 10 days https://t.co/ZSXmzVFqB1

— Elon Musk (@elonmusk) September 2, 2026

SpaceX Data Sits at the Center of Grok 4.7

Musk has built the case across a run of posts. He said in mid-August that initial training had finished and that SpaceX company data was being incorporated through supplemental training. 

Musk had earlier detailed the architecture. Grok 4.7 runs on a 2.1 trillion parameter base, up from the 1.5 trillion parameter foundation behind Grok 4.6. He said the larger model runs slightly more slowly while using tokens more efficiently.

He also named the rival he expects to trail him.

“Grok 4.7 will exceed all current models. That said, Anthropic is a great company and will probably release improved models soon. However, the SpaceX training corpus is so awesome & unique that I would be shocked if any model is better at real-world engineering than 4.7,” the post read.

Meanwhile, the release cadence has tightened. SpaceXAI took Grok 4.5 public in July and shipped Grok 4.6 on August 12. 

Grok 4.6 Benchmarks Set the Bar

Grok 4.6 was built on its predecessor, Grok 4.5. The company said the model has a “particular focus on long-running agents and more ambitious interactive and visual work.”

According to figures shared, the model scored 61 on the AA Intelligence Index, level with GPT-5.6 Sol Max and behind Claude Fable 5 Max at 62. Grok 4.5 scored 56.

Introducing Grok 4.6.It delivers frontier intelligence and is a significant improvement over Grok 4.5 at the same price. pic.twitter.com/RtTbpXcb3a

— SpaceXAI (@SpaceXAI) August 12, 2026

Grok 4.6 led GDPVal-AA v2 with 1753. Yet it managed 26% on Terminal-Bench v3.0, well behind GPT-5.6 Sol Max at 34.6%.

Grok 4.5 set a similar pattern. It topped Artificial Analysis’s AutomationBench-AA at 51.4% while costing $0.34 per task. However, it logged 0.63 guardrail violations per task, above Claude Opus 4.8’s 0.55.

Musk has now attached a firm number to the release. Whether SpaceXAI ships independent evaluations will determine how far the engineering claim travels.

Subscribe to our YouTube channel to watch leaders and journalists provide expert insights