SpaceXAI announced Grok 4.7 on September 21 as its latest model for coding and general knowledge work. The company says the release is meant for tasks that take longer to plan and execute, while keeping the price and speed of Grok 4.6.
According to SpaceXAI, 4.7 uses a larger base model and a longer reinforcement-learning run than its predecessor. The company says that training emphasized harder problems that can take hours to complete, and that the model is better at checking its own work and managing long context. Those are company performance claims, not independent measurements.
Where the model is available
SpaceXAI says Grok 4.7 is available through its API, third-party coding harnesses, model routers and cloud platforms. Cursor's documentation lists it as a first-party model for its desktop app, web and cloud agents, CLI, SDK, automations and iOS app, with availability depending on a user's plan and a team's model settings.
Cursor documents a 256,000-token standard context window and a 500,000-token long-context option. That matters for the audience the release is chasing: coding and research work that needs a model to retain a large codebase or a long chain of material without reducing the task to a short chat exchange.
The announcement is a product rollout, not an independent comparison. SpaceXAI points to its own benchmark results and price-performance positioning, while users will still need to judge the practical trade-offs in their own tools, workloads and access plans.
