xAI moved Grok 4 to general availability today, ending the beta period that began last fall. All X Premium subscribers now get access; X Premium+ ($16/month) gets priority routing and higher rate limits. The standout feature is a per-message reasoning toggle — a slider that controls how many internal CoT tokens the model uses before responding, ranging from "fast" (no reasoning) to "deep" (multi-minute thinks).
The distribution play matters. xAI inherits a captive audience of paying X users who can flip Grok on without signing up for yet another AI subscription. That's structurally different from Anthropic and OpenAI, which both rely on standalone consumer apps for non-API revenue. The question is whether Grok 4 quality is good enough to convert casually-curious X users into engaged AI tool users.
On benchmarks Grok 4 lands roughly between Claude Sonnet 4.5 and Gemini 2.5 Pro on most evals — competitive but not category-leading. Where it stands out is real-time information from the X platform itself; Grok has a structural advantage on "what's happening on the internet right now" queries that the closed-platform competitors can't match.