Musk Ships Grok 4.5 and Calls It Opus-Class

SpaceXAI released Grok 4.5 to the public on Thursday, one day after Elon Musk announced it and described the model as "an Opus-class model, but faster, more token-efficient and lower cost." His internal read, he said, puts Grok 4.5 "roughly comparable to Opus 4.7, but much faster."

The claim is worth reading carefully. "Opus-class" is Musk's framing, not an independent finding, and it rests on the company's own assessment rather than a public benchmark run. What is verifiable is the price. SpaceXAI is charging $2 per million input tokens and $6 per million output, against the $5 and $25 that Anthropic lists for Opus 4.7. If the quality claim holds up under outside testing, the pricing is aggressive. That is the part still waiting for evidence.

Built on Cursor's data

Grok 4.5 sits on what the company calls a 1.5-trillion-parameter V9 foundation, and it was trained in part on data from the AI code editor Cursor, part of a partnership that also makes the model available inside Cursor on every plan. It is reachable through Grok Build and the SpaceXAI console as well, though not yet in the European Union.

The coding focus is deliberate. The most valuable and most contested ground in the model business right now is agentic software work, where a model runs for long stretches, calls tools, and fixes its own mistakes. Training on a real editor's usage data is a bet that Grok can compete there rather than on general chat.

A model a month

The launch lands a day after the company completed its own reinvention. SpaceXAI is the entity that emerged when Musk folded xAI into SpaceX earlier this week, and Grok 4.5 is its first model under the new name. Management has said it intends to release a new foundation model every month through the end of 2026, a cadence no other major lab has committed to publicly.

That pace is the real story to watch. Monthly foundation models, if SpaceXAI can actually sustain them, would put unusual pressure on rivals who ship on a slower rhythm. It also raises an obvious question about how much genuine change each release can carry. Shipping often is easy to announce and hard to do well. Grok 4.5 arrived on the same day OpenAI opened GPT-5.6 to the public, so buyers now have two fresh, cheaper flagships to test against each other, and a clearer sense of where the price of frontier work is heading.