Grok 4.6 reaches AWS, Google and Microsoft clouds within two weeks

SpaceXAI released Grok 4.6 on August 12, 2026 with a 500K-token context window, configurable reasoning effort and a focus on long-running agents, coding and knowledge work. Within two weeks the model became available through Amazon Bedrock on August 19, Google’s Gemini Enterprise Agent Platform on August 21 and Microsoft Foundry on August 26. AWS exposes Grok 4.6 through Bedrock with cross-Region inference and US-only processing options; Google exposes it through Model Garden; Microsoft exposes it through the Foundry model catalog. SpaceXAI lists base pricing of $2 per million input tokens, $0.50 per million cached-input tokens and $6 per million output tokens on the AWS and Google integrations.

The material development is distribution rather than another benchmark claim. Teams already standardized on AWS, Google Cloud or Microsoft’s enterprise AI stack can now test Grok 4.6 without building a separate direct-provider integration and can retain the cloud platform’s existing model catalog, access controls, logging, billing and—in AWS’s case—explicit regional-processing profiles. That lowers the operational cost of multi-model evaluation and makes model switching easier for long-running agent workloads. It also increases price and capability competition inside the managed-model catalogs where many enterprise builders already make deployment decisions.

Grok 4.6 launched on August 12 through SpaceXAI’s own API and partners including Cursor, OpenRouter, Vercel and Cloudflare. Amazon Bedrock added it on August 19, Google’s Gemini Enterprise Agent Platform followed on August 21, and Microsoft Foundry added it on August 26. The important change is that the same model can now sit inside several existing enterprise procurement and governance environments rather than requiring a new standalone vendor path.

Amazon Bedrock supports both Global and US Geo cross-Region inference profiles for Grok 4.6. AWS says the US profile keeps processing within the United States while the global profile can route across supported commercial regions for greater capacity and lower per-token cost. Bedrock also exposes the model through Responses, Chat Completions and Converse APIs and keeps normal Bedrock controls such as invocation logging, CloudWatch metrics and Cost Explorer reporting.

Google exposes Grok 4.6 through Model Garden in Gemini Enterprise Agent Platform, while Microsoft Foundry places it in the same catalog used to evaluate and deploy other frontier models. That matters for organizations that want workload-specific model selection without building a separate operational control plane for each provider.

SpaceXAI positions Grok 4.6 for long-running agents, coding and interactive work and publishes strong benchmark results, but most headline numbers are provider-reported or harness-sensitive. The new cloud availability makes independent comparison easier; it does not establish that Grok 4.6 is the best choice for any particular workload. Builders should compare task success, latency, token consumption, tool behavior and governance requirements in the cloud environment they actually use.