SpaceXAI has released Grok 4.7, its latest flagship AI model targeted at coding, agentic tasks, and knowledge work. According to reporting by MarkTechPost, the new model is built upon a larger base model and underwent an extended reinforcement learning run compared to its predecessor. Despite these architectural changes, SpaceXAI is keeping the pricing identical to the previous version, maintaining the two-dollar and six-dollar pricing structure established with Grok 4.6. The company confirmed that the model retains the same operating speed as Grok 4.6 and is currently available for deployment as a hosted model through the standard api call designation.
Pricing and Deployment Details
The decision to maintain flat pricing while increasing the underlying base model size and extending reinforcement learning training points to ongoing efficiency gains in model scaling and inference operations. For engineering teams and enterprise developers evaluating hosted infrastructure options, cost predictability is a critical factor when migrating workloads between model generations. By preserving the exact pricing tier of the 4.6 release while expanding capability parameters for complex programming workflows and multi-step agent actions, SpaceXAI is attempting to capture a larger share of production-grade workloads without increasing operational budgets for end users.
What Remains Unknown
Despite the release, several technical specifications have not yet been disclosed. The announcement did not specify the exact parameter count of the larger base model, nor did it detail the precise duration or computational scale of the reinforcement learning run. Furthermore, public benchmark scores comparing Grok 4.7 against competing foundational models in specific software engineering or logical reasoning tasks were not provided in the initial rollout details. Observers will need to monitor independent evaluations and upcoming technical documentation to fully assess performance shifts relative to Grok 4.6.
Further updates regarding performance benchmarks and extended API availability are expected as developers begin integrating the hosted model into production environments.
Lead image: panumas nikhomkhai / Pexels
