A multimodal model built for long-running agents, with selectable reasoning effort, now also on Microsoft Foundry and Amazon Bedrock.
Grok 4.6 launched on August 12 as a multimodal model handling text, images, and audio, with a 500k-token context window and four reasoning-effort settings: low, medium, high, and xhigh.
xAI pitched it at long-running agents, the same kind of work Grok Bot does, and made it available beyond its own API on Microsoft Foundry and Amazon Bedrock.