Friday, Oct 30 | --:--
Back to home

Grok 4.6 Completes the Big-Three Cloud Map on Microsoft Foundry

xAI put Grok 4.6 on Microsoft Foundry in public preview at $2 input / $0.50 cached / $6 output per million tokens (Global Standard) — two weeks after the model GA and a week after Amazon and Google cloud ports. Watch: Foundry Learn docs list 200k context against xAI’s 500k card.

Times of AI Desk 3 min read San Francisco, CA View as Markdown
Engraved Grok mark nested in a Microsoft Foundry catalog plate marked 4.6, security-intaglio steel-blue linework. No currency-note layout.

Corrected October 14, 2026: Sourced Grok 4.6's Foundry price to Microsoft's blog and removed an unverified claim that it matches the xAI API and Bedrock prices.

Grok 4.6 is now on AWS, Google, and Microsoft inside two weeks. That is a procurement story as much as a model story: enterprises can keep IAM, VPC, and existing cloud commits and still call the new flagship.

xAI said Grok 4.6 is now available via Microsoft Foundry. Microsoft’s Foundry blog (Mahesh Balachandran, same day) calls it public preview in the Foundry Models catalog: evaluate against other frontier models, deploy managed endpoints, enterprise governance. Microsoft lists Global Standard pricing of $2 input / $0.50 cached / $6 output per million tokens. xAI states 500k context (Microsoft's catalog lists 200k) and low / medium / high / xhigh reasoning. Microsoft Learn later documents grok-4.6 (preview) as 200k context, Global Standard, Chat Completions + Responses, model-format "xAI" — which does not match xAI's 500k figure. @Azure posted the catalog link the same afternoon.

What shipped#

LinkEmbed

Item x.ai/news + Microsoft Foundry blog Aug 26
Surface Microsoft Foundry Models, public preview
Price $2 / $0.50 cached / $6 per 1M tokens (Global Standard)
Positioning Long-running agents, coding, engineering, office, research

Grok models were already on Azure/Foundry in older SKUs (grok-4, grok-4.1-fast-reasoning). This post is 4.6 specifically.

Same cloud-distribution pattern as OpenAI Daybreak-on-Bedrock.

Limits#

  • Foundry’s 200k context in Learn may be a preview cap or a docs lag on xAI’s 500k card — unresolved here.
  • Public preview, not GA on Foundry.
  • Microsoft’s “1.5T-scale” framing is partner marketing, not an independent parameter count.

Sources#

Prior Coverage

Earlier Times of AI reporting on this thread.

Scroll to continue reading