Elon Musk said SpaceXAI will release Grok 4.6 around Aug. 7 and follow it weeks later with Grok 4.7, a larger 2.1 trillion parameter model. Key Points: Musk put Grok 4.6 on a target of roughl
Elon Musk said SpaceXAI will release Grok 4.6 around Aug. 7 and follow it weeks later with Grok 4.7, a larger 2.1 trillion parameter model.
Key Points:
- Musk put Grok 4.6 on a target of roughly Aug. 7, with Grok 4.7 a few weeks behind it.
- He described the follow-up as a 2.1 trillion parameter model that answers more slowly but uses fewer tokens.
- The schedule arrives days after Kimi K3 weights went public and Anthropic shipped Claude Opus 5.
Musk laid out the schedule on X on Tuesday, describing Grok 4.6 as a build carrying substantially heavier supervised fine-tuning and reinforcement learning than the model it replaces.
He put its arrival at roughly Aug. 7. Grok 4.7 would land a few weeks behind it, and he framed the larger system as an upgrade on every measure except the speed at which it answers.
He argued the slower model would still finish the same work with fewer tokens, which is the efficiency case SpaceXAI has pressed since the spring.
The 2.1 trillion figure also sits awkwardly beside his own earlier account, which pegged Grok 4.6 at 2 trillion parameters when he confirmed its training timetable on Jul. 18.
Also Read:Crypto Wrench Attacks Reach 52 Cases In Six Months, Most Of Them In France
Timing matters here. Moonshot AI released the full weights for Kimi K3, a 2.8 trillion parameter model, on Monday, and Anthropic shipped Claude Opus 5 three days before that. Artificial Analysisscored Kimi K3 at 57 on its intelligence index against 54 for Grok 4.5, though the smaller Grok model closed the same tasks for roughly $0.31 apiece versus $0.94.
Guillermo Rauch, chief executive of Vercel, said his company's newest internal tests rank Grok 4.5 first on cybersecurity price-performance, landing close to Kimi K3 on results while running several times cheaper than Claude Opus 5. He put OpenAI's Sol model at the frontier regardless, still ahead of the Anthropic release.
That gap between price and rank is the whole SpaceXAI pitch, and Grok 4.5 posted roughly a quarter of the output tokens Claude Opus 4.8 burned through to clear the same set of software engineering problems.
Tuesday's post caps a month of shifting targets. SpaceXAI shipped Grok 4.5 on Jul. 8 at $2 per million input tokens and $6 per million output, less than half what Anthropic charges for Opus 4.8.
Musk called the model roughly comparable to a year-old Claude flagship, then told followers on Jul. 24 that Grok 4.6 was two weeks away and Grok 4.7 four.
Each of those claims arrived on X rather than in a model card, and the company has published no specifications, pricing, context window or benchmark results for either unreleased model. Musk has missed his own stated AI deadlines before, and neither release has an official launch page yet.
Read Next:XRP Near-Term Forecasts Top Out At $1.25 Before The Fed Decision
