SpaceX launches Grok 4.7 at $4.69 per task, beating rivals

Grok 4.7 arrives less than a week after SpaceX's previous model release, a text-to-speech model introduced on Friday. That model, Grok Voice Transcribe 2.0, offered double the accuracy of its predecessor at half the cost and, according to SpaceX, outperformed several competitors' text-to-speech offerings. The rapid cadence suggests SpaceX is pushing to expand its AI portfolio across both language and voice categories.

The Grok line originated with xAI Corp., an Elon Musk-founded startup. Musk merged the company with xAI Inc. last year, and the combined organization was subsequently folded into SpaceX, which gained the Grok model series along with multiple AI data centers. Grok 4.7 also draws on Cursor, another recent SpaceX acquisition, for its CursorBench evaluation.

SpaceX attributes Grok 4.7's performance to a new base model and an enhanced reinforcement learning workflow, where the model received tougher training tasks and worked on them longer than its predecessor. The model is compatible with SpaceX's Grok Bot harness, enabling it to split work among multiple AI agents that run tasks in parallel and verify each other's outputs. Whether that architecture translates into sustained cost or quality advantages across more benchmarks is likely to shape how enterprise buyers weigh Grok 4.7 against alternatives like GPT-6 Astra.