Grok 4.7 claims near-Opus 5 results on certain tasks

The pitch for Grok 4.7 is built around somebody else's model: close to Opus 5 performance on certain tasks, at the price of Grok 4.6. That is how a post from the SpaceXAI account on Threads sums up the release, pointing to a page on x.ai for the rest.

At a glance

• Grok 4.7 is listed as live on Grok Build, in Cursor and through the APIs, according to the SpaceXAI post, which links to a release page at x.ai rather than spelling the details out.
• The selling point is price against performance: results near Anthropic's Opus 5 on some tasks, charged at what the previous model, Grok 4.6, was charged at.
• No benchmark figures travel with that claim, so which tasks come close, by how much and against which Opus 5 setup are all unstated in the post itself.

If you have not been following the naming: Elon Musk has called Grok 4.5 an Opus-class model, so measuring a Grok release against Anthropic's top line is a habit here rather than a one-off. The company behind it has also changed shape. According to note.com, xAI was rebranded or reorganised as SpaceXAI, which is the name on the account making the announcement.

Three places to find the model: Grok Build, Cursor and the APIs. For anyone with work already wired up, the last one matters most, because an API release drops into an existing pipeline without anyone switching editors. The post lists all three together, with no staggering between them.

Price is the other half of the claim, and it is a claim of no change. A version bump normally arrives with a new rate card; this one, SpaceXAI says, arrives at what Grok 4.6 costs. The upgrade is framed as something you take without redoing the budget line for it.

The comparison itself points sideways rather than backwards. The post measures Grok 4.7 against Anthropic's Opus 5, not against the Grok it replaces, and it scopes the comparison to certain tasks rather than to everything. That qualifier is in the sentence doing the selling, which is worth noticing before you read the headline number as a general one.

There is no number. Which tasks come close to Opus 5, by how much, and under which configuration are not stated, and neither is the token price that parity with Grok 4.6 actually works out to. Leading with a competitor's model and then shipping the claim without figures is, in our view, an odd way to sell the cheaper option.

Where the numbers should land The post signs off with a line about testing, and that is the part you can act on yourself. Until someone outside the company runs the comparison, near-Opus 5 stays a claim rather than a result. Worth watching for: the first side-by-side runs from people already using it in Cursor, and a task list specific enough to check a second time. Nothing in the post points to a further rollout beyond the three surfaces already named.