SpaceXAI launches Grok 4.7, its "most powerful model for coding and knowledge work"

Grok 4.7 has been officially launched, and with this new model, SpaceXAI is going after the growing market for AI models built for coding and serious knowledge work. The new model uses a larger base model than Grok 4.6, runs through a longer reinforcement learning process, and is designed to spend more time working through difficult tasks before producing an answer.

SpaceXAI says Grok 4.7 is twice as fast at half the price of comparable models, while keeping the same listed pricing and speed as Grok 4.6. The model is available starting today through Grok Build, Cursor and the Grok API, along with third-party coding harnesses, model routers and cloud platforms.

Grok 4.7 is built for longer and harder tasks

SpaceXAI says Grok 4.7 was trained on a harder mix of tasks weighted toward problems that can take several hours to complete. The model is also better at checking its own work, managing longer context and handling multi-step tasks. The co.pany also mentions that it was trained to understand the Grok Bot harness natively, which should also improve conversational tasks and broader knowledge work.

On CursorBench 4.0, Grok 4.7 scores 46.3%, compared with 40.4% for Grok 4.6. It also reaches 71.0% on DeepSWE v1.1 using high effort, while scoring 64.0% on EEBench for electrical engineering.

SpaceXAI is also highlighting professional work as it mentions Grok 4.7 reaches 1,735 Elo on GDPval and improves over Grok 4.6 on AA Briefcase, a benchmark focused on multi-hour office tasks.

Grok 4.7 price and availability

Grok 4.7 starts at $2 per million input tokens and $6 per million output tokens. xAI also offers a faster variant with twice the output speed at twice the price. The model includes a new safeguard stack that xAI says improves jailbreak resistance and dangerous-task refusals. On LatchBio’s biosafety benchmark, Grok 4.7 scores 62.4%, while xAI says only 3.3% of risky dual-use cybersecurity prompts were allowed through on HackerBench v0.3.