Quick answer: Microsoft says Grok 4.7 is now generally available in Microsoft Foundry Models. The model is positioned for advanced reasoning, software engineering, long-horizon tasks and agentic workflows. Foundry supports Global Standard and US Data Zone Standard deployments, Chat Completions and Responses APIs, streaming, tool calling, and authentication through Microsoft Entra ID or API keys.
The important detail for enterprise buyers is that Foundry pricing varies by context length and deployment geography, so teams should not assume the direct-provider rate applies inside Azure.
Grok 4.7 Microsoft Foundry pricing
Microsoft’s launch post lists separate pricing bands for short and long context in the US Data Zone. The published table shows:
| US Data Zone | Input | Output | Cached input |
|---|---|---|---|
| Short context under 200K | $2.20 / 1M tokens | $6.60 / 1M tokens | $0.55 / 1M tokens |
| Long context 200K+ | $4.40 / 1M tokens | $13.20 / 1M tokens | $1.10 / 1M tokens |
Pricing can change and other deployment types can differ, so production teams should verify the current Foundry model card before forecasting spend.
How much context does Grok 4.7 support?
Microsoft says Grok 4.7 is planned to support up to 500K tokens of context from launch. That is particularly relevant for large repositories, long legal or technical documents, multi-file analysis and agent workflows that need to retain more state than a short chat session.
Available deployment options
- Global Standard: designed for broad availability and flexible capacity.
- US Data Zone Standard: useful for customers that need US data-zone processing constraints.
The exact region availability and quotas can depend on subscription, model access and capacity. Enterprise teams should confirm deployment eligibility in their own Foundry project instead of assuming every Azure region exposes the same model catalog.
API support
Microsoft says developers can integrate Grok models with familiar Chat Completions and Responses APIs. The Foundry integration also supports streaming and tool calling, which makes it suitable for interactive assistants and multi-step agents.
Authentication choices
Teams can use Microsoft Entra ID or API keys. For production environments, Entra ID is attractive because it can fit existing identity, role and secret-management controls. API keys remain simpler for prototypes and isolated services.
What workloads is Grok 4.7 aimed at?
Microsoft highlights advanced engineering, reasoning, planning and execution. That makes the model a candidate for:
- Large codebase analysis and software-engineering agents.
- Technical research across long documents.
- Tool-using enterprise agents.
- Long-running workflows that must preserve context.
- Multi-model routing where teams compare several frontier models inside one platform.
Why Foundry access matters
For Azure-heavy organizations, the main benefit is not simply another model endpoint. Foundry provides a common place to discover, evaluate, deploy and govern models. Teams can compare Grok 4.7 with other providers without building a separate governance stack for every vendor.
This mirrors the broader multi-model trend AVARIXO covered with GLM 5.3 on Amazon Bedrock: cloud platforms increasingly compete on model choice, identity, observability and deployment controls rather than a single exclusive model.
Short context vs long context: what changes?
The pricing split at 200K tokens means prompt design matters. A workflow that repeatedly sends a 250K-token repository snapshot can cost materially more than one that uses retrieval, caching, chunking and smaller context windows. Cached-input pricing can reduce the cost of stable prefixes, but architects still need to measure real request shapes.
What to test before production
- Context length under your real documents, not synthetic prompts.
- Tool-calling accuracy for your schema.
- Latency at long context.
- Cached-input hit rate.
- Regional availability and quota.
- Entra ID permissions and managed identity flow.
- Output quality against the models already in your routing layer.
Sources
FAQ
Is Grok 4.7 generally available in Microsoft Foundry?
Yes. Microsoft’s launch post describes the model as generally available in Microsoft Foundry Models.
Can I use Entra ID instead of an API key?
Yes. Microsoft says the integration supports both Microsoft Entra ID and API-key authentication.
Does Grok 4.7 support 500K context?
Microsoft says support is planned up to 500K tokens from launch. Check the current model card for the limit available to your deployment.
Other Articles
