SpaceXAI introduced Grok 4.7 on September 21, 2026 as its most capable model for coding and knowledge work. The company says the model works longer on difficult tasks, verifies its own output more carefully, and manages longer context better than Grok 4.6. It also says Grok 4.7 was trained to understand the Grok Bot harness, improving conversational and general knowledge-work tasks.
The article opens by identifying Grok 4.7 as SpaceXAI’s newest model for coding and knowledge work; the official release artwork directly identifies that subject.
The clearest technical change is the model’s training emphasis. SpaceXAI says it used a larger base model and a longer reinforcement-learning run on harder tasks, especially problems that can take hours to complete. That matters because many AI coding evaluations reward quick fixes, while real development often requires sustained work, context management, and repeated self-checking.
The supplied results show meaningful gains over Grok 4.6 in several areas. SpaceXAI reports scores of 46.3% versus 40.4% on CursorBench 4.0, 71.0% versus 65.2% on DeepSWE v1.1, 64.0% versus 53.0% on EEBench, and 1,657 versus 1,546 on AA Briefcase v1.1. The DeepSWE result is marked as a high-effort score, so it should not be read as a like-for-like default-setting comparison.
The pricing claim is equally specific: SpaceXAI lists Grok 4.7 at $2 per million input tokens and $6 per million output tokens, the same figures it gives for Grok 4.6, and says the model is served at the same price and speed. It also says a faster variant is available at twice the output speed and twice the price. Those are publisher-stated commercial terms, not independently verified market measurements.
The evidence does not support a blanket claim that Grok 4.7 beats all frontier models. In the supplied table, Fable 5.1 scores higher on CursorBench, GPT-5.6 Sol scores higher on DeepSWE, and GPT-6 Astra scores higher on Terminal-Bench and HealthBench Professional. The packet also lacks independent testing, so the most defensible conclusion is narrower: SpaceXAI presents Grok 4.7 as a stronger and more efficient Grok 4.6 successor, with notable gains in selected long-running and professional-work benchmarks.