xAI Releases Grok 4.6 for Long-Running Agents and Expands It to Bedrock
xAI released Grok 4.6 with a 500K context window and a focus on long-running agents, then made the flagship model generally available through Amazon Bedrock.
Grok 4.6 is released
xAI released Grok 4.6 on August 12, 2026, describing it as an update focused on long-running agents, complex coding, knowledge work and more ambitious interactive and visual tasks. The model is available through xAI's own products and API, with subsequent distribution through major developer platforms.
The company says Grok 4.6 is designed to stay on task across many steps while researching, analyzing information, working across codebases and producing applications or other work artifacts. xAI reports frontier-level results across several agentic coding and knowledge-work evaluations, but those benchmark claims should be read in the context of the developers' published harnesses and evaluation conditions.
500K context and configurable reasoning
For Amazon Bedrock, xAI lists a 500,000-token context window and configurable reasoning effort settings of low, medium, high and xhigh. That combination targets long-horizon workflows where an agent needs to carry large codebases, documents, intermediate tool results and persistent task state across multiple steps.
The model's published Bedrock pricing is $2 per million input tokens, $0.50 per million cached input tokens and $6 per million output tokens. Pricing and regional availability can change, so production users should confirm current provider documentation before estimating workloads.
General availability on Amazon Bedrock
On August 19, xAI announced that Grok 4.6 became generally available on Amazon Bedrock in supported AWS Regions. This gives AWS customers a managed path to use the model through Bedrock controls and infrastructure rather than integrating only with xAI's direct API.
The Bedrock release is distinct from Grok 4.6's original model launch: the model itself was already released, while the August 19 update expands where developers can deploy it. xAI has also been extending Grok 4.6 to other development platforms, reflecting a broader multi-cloud distribution strategy.
What developers should evaluate
The main practical questions are not only benchmark scores but whether Grok 4.6 maintains task coherence over long sessions, uses tools reliably, controls token costs across a 500K context window and integrates cleanly with an organization's existing agent harness. Caching can materially affect long-context economics, while very high reasoning settings can increase cost and latency.
Teams comparing models should test representative repositories and workflows with consistent prompts, tools, time limits and retry policies rather than assuming vendor-reported benchmark rankings transfer directly to production.
Release status
Grok 4.6 is a released model, not an announced future system. Its Amazon Bedrock availability is also generally available according to xAI's August 19 announcement.
