SpaceXAI, Elon Musk's AI company, released Grok 4.6 on August 12, 2026, describing it as a major upgrade of Grok 4.5 built around long-running agents and more ambitious interactive and visual work. The model stays on task across many steps — researching a topic, analyzing information, working across a codebase, or turning an idea into a polished application, the company said.
The headline benchmark claim is cost efficiency: Grok 4.6 scores 61 on the Artificial Analysis Intelligence Index, tying OpenAI's GPT-5.6 Sol (61) and sitting just one point behind Anthropic's Claude Fable 5 (62), the index leader. On the GDPVal-AA v2 knowledge-work benchmark it posted 1753 Elo — the best of the group compared by the company — and it led on CursorBench v3.2 at 69.9%.
The pricing is what makes the release notable: $2 per million input tokens and $6 per million output tokens, with a faster variant at double the price. Rivals charge several times that for comparable frontier intelligence — per Artificial Analysis, the measured cost per Intelligence Index task for Grok 4.6 is about a third of Claude Fable 5's and roughly 60% below comparable frontier models. SpaceXAI is sweetening the launch with 2x included usage inside Grok Build and Cursor for the first week.
On training, SpaceXAI says Grok 4.6 went through a longer supplemental run with curated model-generated data, an improved optimizer, and SFT trajectories regenerated by Grok 4.5 itself, followed by agentic reinforcement learning across domains including kernel optimization, web development and computer-aided design. The company reports stronger first passes on visual and interactive projects and more self-testing and verification on long trajectories.
Grok 4.6 is available today in Cursor and Grok Build, as well as through the SpaceXAI API and partners including OpenRouter, Vercel and Cloudflare. The release continues a crowded month for frontier AI: within the past six weeks, six labs — Anthropic, OpenAI, Moonshot AI, SpaceXAI, Z AI and Meta — have fielded models scoring above 50 on the Intelligence Index, up from just two in early June.
Sources
- x.aiIntroducing Grok 4.6 — SpaceXAI
- venturebeat.comSpaceXAI debuts Grok 4.6, overtaking Kimi K3's performance and matching GPT-5.6 Sol (VentureBeat)
- artificialanalysis.aiGrok 4.6 returns SpaceXAI to the intelligence frontier — Artificial Analysis
- marktechpost.comSpaceXAI Releases Grok 4.6: A 500K-Context Frontier Model (MarkTechPost)




