xAI releases Grok 4.6 with 500K context and launches $300/month Grok Bot cloud agent

xAI released Grok 4.6, extending its Grok model series with a 500K-token context window and coding/agentic capabilities. Pricing is aggressive at $2 per 1M input tokens and $6 per 1M output tokens — roughly $8 per 1M combined, versus $30–60 for Anthropic's Opus and Sonnet tiers. xAI also launched Grok Bot, a $300/month persistent cloud agent service that runs autonomously in the cloud rather than requiring a live session.

The pricing makes Grok 4.6 one of the week's several cost-disruptor releases, alongside Gemini 3.7 Flash and Qwen3.8-Flash-Next. Developers welcomed the value but flagged a long-context billing cliff where costs double at 200K tokens — a friction point for exactly the agent workloads Grok is pitched at, since agentic loops accumulate context quickly. The $300/month Grok Bot subscription targets users who want an always-on agent, competing with emerging persistent-agent offerings.

Grok 4.6 also became available on AWS Bedrock (as 'SpaceXAI's Grok 4.6') with cross-region inference, broadening its enterprise reach beyond xAI's own API. Competitively, xAI is leaning into price-performance to claw share from the premium labs while the market commoditizes mid-tier models amid 'release velocity fatigue.' The persistent-agent angle is the more differentiated bet. Watch whether the 200K billing cliff deters agent developers and how Grok Bot's autonomy compares to AWS AgentCore and other cloud-agent frameworks now reaching GA.