xAI has released Grok 4.6, an update focused on sustaining long-running agentic tasks and producing stronger first attempts at interactive and visual projects, the company announced on 12 August 2026.
xAI said Grok 4.6 matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, a composite of nine benchmarks, with both models scoring 61, narrowly behind the leading model on that index and ahead of the previous Grok 4.5 High, which scored 56. xAI said competitor figures are drawn from developers' own published system cards or benchmark leaderboards rather than independently verified.
The model underwent a longer supplemental training run than Grok 4.5, using curated model-generated data, improved engineering data, and a revised optimiser and training recipe, followed by supervised fine-tuning trajectories regenerated using Grok 4.5 across reasoning efforts and domains including STEM, software engineering and knowledge work. xAI said the model shows more self-testing and verification behaviour on longer tasks, checking its own work before proceeding.
xAI said Grok 4.6's safeguards have been recalibrated in line with its expanded capabilities, with what the company described as its widest-ever suite of pre-deployment safety testing alongside post-deployment and third-party evaluation.
Grok 4.6 is available today through Cursor, Grok Build, the API, and partners including OpenRouter, Vercel and Cloudflare, priced from $2 per million input tokens and $6 per million output tokens, with a faster variant at double that price. xAI is offering double included usage in Grok Build and Cursor for the first week.
