I watched the alert ping across a midnight chat. Engineers in two companies tightened their focus and a model quietly moved from lab to beta. You can feel the pressure building; something big is near.
I report on this so you don’t have to chase threads. I’ll walk you through what matters, who’s involved, and why that $60 billion (€55 billion) Cursor deal suddenly reads like a pivotal move in AI strategy.
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.
RL is continuing to significantly improve the model, and the Grok Build…
— Elon Musk (@elonmusk) June 28, 2026
In a cramped war room, engineers swapped test results and timestamps
SpaceXAI — the rebrand of xAI announced this week — is reportedly preparing to push a major model into the wild as soon as Wednesday, according to The Information. I’m telling you this because the report names Cursor as a co-creator on the new model, the first production built jointly by SpaceXAI and Cursor after the companies moved toward a $60 billion (€55 billion) acquisition.
The practical detail that matters: Musk has already said a May Grok release included Cursor data, and he posted that Grok 4.5 runs on a 1.5T V9 foundation with supplemental Cursor training. Those are not marketing flourishes; they’re technical breadcrumbs pointing to a model being tuned across SpaceX and Tesla environments.
When will SpaceXAI release its new model?
The Information’s timeline says “as early as Wednesday.” I don’t pretend to read internal calendars, but the cadence fits Musk’s recent public hints and private betas. If you’re tracking for competitive or integration reasons, prepare for rapid updates: this is an event announced by leaks, then confirmed by an internal rollout.
At a developer desk, someone ran benchmarks and compared scores
Internally, sources say the new model is being measured against Claude Opus 4.8 and OpenAI’s GPT-5.5—while GPT-5.6 exists and is expected soon. I watched similar measure-ups in past cycles; they’re noisy, political, and useful. Benchmarks aren’t the full story, but they set expectations for adopters and investors.
If you follow model names, note the language: Opus 4.8, GPT-5.5, Grok 4.5. Those decimals matter to teams tuning RL and safety layers. Michael Truell, Cursor’s founder, flagged last month that Cursor had a model meant to compete with Anthropic and OpenAI frontier products—he called it the company’s “next phase,” without naming SpaceXAI directly at the time.
How does Grok 4.5 compare to GPT-5.5 and Claude Opus?
Public claims are mixed. Musk suggested Grok 4.5’s early evaluations were “close to, perhaps exceeding Opus.” That’s a strong claim from a founder; you should treat it like a market signal rather than proof. From where I sit, the likely reality is competitive parity in some benchmarks and domain advantages where Cursor-supplied data aligned with SpaceX/Tesla telemetry and developer needs.
On a trading floor, analysts re-priced bets on platform strategy
The $60 billion (€55 billion) acquisition chatter has traders and execs reshuffling assumptions about speed-to-market and access to proprietary datasets. If Cursor brings unique conversational fine-tuning or tooling, SpaceXAI gains more than model weights — it gains a product path into developer workflows and potentially enterprise contracts.
Think of this moment like a pressure valve about to hiss: the leak of timing plus Musk’s tweet forces competitors to react. You should also picture the partnership as a chess grandmaster sweeping a board — moves made now will shape where others play next.
There are open questions that matter to you and to any team planning integrations: how rigorous are the safety controls, what datasets shifted performance, and will models be released to external devs or kept behind SpaceX/Tesla gates? Look to Anthropic, OpenAI, and recent testbed outputs for clues, but expect fast iterations and public sparring.
Who benefits? Developers building specialty copilots, enterprises wanting alternative supply chains to OpenAI, and speculative investors who prize differentiation. Who loses time? Companies that wait for a clear winner before testing migrations.
I’ll keep watching the signals — press releases, follow-up leaks, and what the private betas leak into GitHub and API decks. Are you ready to update your assumptions when the next benchmark drops?