SpaceXAI has released Grok 4.6, its latest AI model focused on long-running agents, coding, knowledge work, and more complex visual projects, while pricing the model well below several competing frontier systems.
The company says Grok 4.6 can stay focused across longer tasks, work through unfamiliar codebases, research technical topics, and test more of its own output before moving forward. SpaceXAI also says the model produces stronger first attempts when users ask it to build interactive applications or visual projects.
Don’t miss the best of The Mac Observer
Set us as a preferred source and our Apple reporting ranks higher in your Google Search results and Discover feed — one tap, no account changes.
Grok 4.6 Matches GPT-5.6 Sol on a Major Benchmark
SpaceXAI says Grok 4.6 scored 61 on the Artificial Analysis Intelligence Index, matching GPT-5.6 Sol at its maximum reasoning setting and finishing one point behind Anthropic’s Fable 5 Max, which scored 62.
Grok 4.6 also improved over Grok 4.5 across every benchmark listed by SpaceXAI, including CursorBench, FrontierCode, APEX-Agents, Terminal-Bench, and AA-Briefcase. The company credits the gains to a longer supplemental training run, higher-quality engineering data, improved optimization, and expanded reinforcement learning across coding and knowledge-work tasks.
SpaceXAI says Grok 4.6 is available through Cursor, Grok Build, its API, OpenRouter, Vercel, and Cloudflare. API pricing starts at $2 per million input tokens and $6 per million output tokens, while the faster version costs twice as much.
Cursor and Grok Build users also receive twice their included Grok 4.6 usage during the model’s first week, giving developers more room to test its agent and coding capabilities.
Discussion