SpaceXAI has released Grok 4.6, a post-training upgrade to Grok 4.5 that features a 500,000-token context window and enhanced capabilities for long-running agents, coding, and knowledge work.
- The model scores 61 on the Artificial Analysis Intelligence Index, tying with GPT-5.6 Sol Max, and leads on GDPval-AA v2 and AA-Briefcase benchmarks.
- It introduces a new `xhigh` reasoning-effort level and accepts text and image inputs with text-only output.
- SpaceXAI reports improved self-testing and verification behaviors in longer trajectories due to supplemental training and reinforcement learning in agentic environments.
- The model is available via the xAI API, Grok Build, and Cursor, with pricing set at $2/$0.50/$6 per 1M tokens for inputs below 200K prompt tokens.
The release targets software engineering, financial research, and legal analysis, offering immediate adoption for seed-stage teams and mid-market engineering organizations through existing integrations.