References to Grok 4.5 have begun surfacing across the products now grouped under SpaceXAI, the banner Elon Musk adopted after folding xAI into SpaceX earlier this year. Two early signals point toward a public launch that may not be far off.
We are now @SpaceXAI. pic.twitter.com/ema66xDWC9
— SpaceXAI (@SpaceXAI) July 6, 2026
The first is a feature flag in the Grok web client that swaps alternate copy into the account upgrade page, where Grok 4.5 looks set to be named; it sits dormant for now and promises nothing about timing, but the scaffolding is in place. The second is the model's appearance in Grok Build CLI, the company's terminal coding agent, within the same flow that prompts users to move up to the Heavy plan before trying it.
🚨BREAKING🚨 Its happening!. Grok Build is showing that Grok 4.5 is Here and prompting users to upgrade to SuperGrok Heavy to access it!
— ️️️️ ️ᅠ️️️️ ️ᅠ️️️️ ️️️️️ ️ᅠ (@blankspeaker) July 6, 2026
You may want to do that quickly to acces the $99/mo promo that is still active right now! https://t.co/FHdmi4TOKH pic.twitter.com/WepllBkEzI
Grok 4.5 is expected to run on a fresh V9 foundation with roughly 1.5 trillion parameters, making it the largest model the company has shipped. Musk has described its performance as close to Opus-level without specifying which version of Anthropic's flagship he meant. That framing carries weight, because the claim rests on internal evaluations at SpaceX and Tesla, where the model has reportedly been running for a couple of weeks, rather than any public benchmark. Independent numbers do not yet exist.
Grok 4.5, based on our 1.5T V9 foundation model, with Cursor data added in supplemental training, is now in private beta at SpaceX & Tesla. Early evals show performance close to, perhaps exceeding Opus.
— Elon Musk (@elonmusk) June 28, 2026
RL is continuing to significantly improve the model, and the Grok Build…
Access looks set to follow the pattern of earlier releases and of grok-build itself:
- Heavy subscribers first
- Then a wider opening to SuperGrok tiers
For SpaceXAI, the model is a test of whether the compute and capital gained through the merger translate into a system that narrows the gap with Anthropic and OpenAI in reasoning and coding, rather than on parameter count alone. The dormant upgrade copy suggests the commercial pieces are being readied ahead of any performance reveal. Whether the release lands on time, against a target that already slipped past late spring, is the open question.