The public preview gives Azure customers access to Grok 4.6 with a 500,000-token context window, extending xAI's multi-cloud distribution push.
By RuntimeWire Staff · Published
Primary source: x.ai Newsroom
Why it matters #
xAI is using distribution as a competitive weapon for Grok. Microsoft Foundry lowers procurement and integration friction, while the public-preview status leaves production maturity for customers to test.
xAI put Grok 4.6 into Microsoft's Foundry model catalog on August 26, giving Azure customers a managed route to the flagship model two weeks after its August 12 release.
The Microsoft Foundry launch is a public preview, according to Microsoft's announcement. Grok 4.6 arrives with a 500,000-token context window and four reasoning settings: low, medium, high and xhigh. xAI is pitching it for coding agents, engineering copilots, research assistants and enterprise automation that must keep working across long sequences of steps.
Musk co-founded xAI in March 2023 around a stated mission to understand the universe. After SpaceX announced its acquisition of xAI, his immediate commercial task is more ordinary: getting Grok into the procurement systems developers and large companies already use.
Grok 4.6 reached GitHub Copilot on August 14, Amazon Bedrock on August 19 and Google's Gemini Enterprise Agent Platform on August 21. RuntimeWire reported on the Bedrock release one week after the model launched. Microsoft Foundry adds Grok to the third major US cloud platform named in xAI's rollout.
Musk's cloud strategy avoids picking a cloud
SpaceX announced its acquisition of xAI on February 2, 2026. Musk is still selling Grok through Amazon, Google and Microsoft. Owning the underlying compute has not changed the need to meet enterprise customers inside their existing cloud accounts.
Microsoft Foundry gives xAI access to customers that may have no interest in establishing another direct AI vendor relationship. The platform offers workload-specific testing, managed endpoints, and enterprise security and governance controls, according to xAI's Microsoft Foundry announcement.
Foundry also makes Grok easier to compare and replace. The platform lets customers evaluate models on their own data and choose among deployment options. Microsoft's model materials place Grok 4.6 beside models from OpenAI, Anthropic, Meta, Mistral, DeepSeek, Cohere and Hugging Face, giving Azure customers one place to compare providers against their own workloads.
For xAI, the trade is worthwhile. Grok gains distribution and a place in enterprise evaluations. Microsoft can benefit from making more model evaluations and deployments available through Azure.
What Azure customers get
Microsoft's public-preview listing gives Azure customers a managed way to evaluate and deploy Grok 4.6 alongside other models. The separate Grok 4.6 announcement lists API pricing starting at $2 per million input tokens and $6 per million output tokens. The supplied announcements do not establish rate limits, regional availability or service-level commitments for the Foundry preview.
The public-preview label matters. It separates this release from the generally available Bedrock deployment xAI announced on August 19. Foundry customers can evaluate and deploy Grok 4.6 through managed endpoints, but the preview status carries less production certainty than a generally available service with established support commitments.
xAI says Grok 4.6 was trained for sustained agent work, including navigating codebases, researching unfamiliar subjects, building applications and checking its own output. The August 12 model announcement attributes the improvement over Grok 4.5 to a longer supplemental training run, curated model-generated reasoning data, engineering data and reinforcement learning across coding, knowledge work, computer-aided design and web development.
Those are xAI's claims. Microsoft notes that the benchmark results in its Foundry announcement come from the model provider, and neither announcement supplies independent Azure customer results. Foundry's practical advantage is that customers can test Grok against their own repositories, documents and workflows instead of treating xAI's launch chart as a purchasing decision.
Distribution becomes part of the model release
Musk's team released Grok 4.6 on August 12 through xAI's API, Cursor and Grok Build. Within 14 days, xAI had added GitHub Copilot and managed access through all three major US cloud platforms.
That pace shows how frontier-model competition has moved into distribution. Raw model capability still matters, and enterprise adoption also depends on billing, identity controls, governance, regional processing and integration with software already approved by security teams. Foundry puts those layers around Grok without requiring xAI to build each one from scratch.
The arrangement also places Grok directly beside the competitors Musk built xAI to challenge. Azure customers can run the same workload through Grok, OpenAI, Anthropic or another provider and compare cost, latency and task completion inside one purchasing environment. xAI gets a chance to win those tests.
For Musk, that is the practical value of the Foundry listing. Grok 4.6 no longer depends on developers seeking out xAI's own console. It can appear in an Azure model picker, under an existing corporate account, ready for the evaluation that determines whether a model becomes part of a production system or another name in a crowded catalog.