SpaceXAI released Grok 4.7 for coding and knowledge work through the Grok API, Grok Build, Cursor, and third-party tools. The model uses a larger base model, and its reinforcement-learning phase gives more weight to tasks lasting hours. The company also says it improved long-context management and self-checking. Published performance results remain company measurements.
SpaceXAI released Grok 4.7 for coding and knowledge work through the Grok API, Grok Build, and Cursor; third-party coding tools and model-routing services can also provide access. The announcement therefore includes an operating release, not just a model preview: developers can call the same version through the API while users select it inside coding environments. IT之家 independently reports the same distribution channels and focus on knowledge work.[1], [2]
The new release uses a larger base model than Grok 4.6. SpaceXAI says it extended reinforcement learning and gave more weight to complex tasks that can last for hours. The model was also tuned to manage long context and check its own output. These changes target coding and research work made up of many steps rather than a single response; both reports present them as improvements described in the company's release materials.[1], [2]
SpaceXAI reports that Grok 4.7 improves on its predecessor across coding, professional knowledge work, and safety evaluations. Those results come from benchmarks selected by the company and its own evaluation setups; the reports do not provide an independent comparison. The model is also described as improved for document and presentation generation and conversational tasks. Its release channels are confirmed, while comparative performance claims remain company statements pending outside measurement.[1], [2]
