Type something to search...
Grok 4.6 available in Microsoft Foundry

Grok 4.6 available in Microsoft Foundry


Introduction

Grok 4.6, the latest frontier model from SpaceXAI, has been available in public preview in Microsoft Foundry Models since August 26, 2026. Built on SpaceXAI’s model family at the 1.5 trillion parameter scale, it does not aim to stand out on an isolated benchmark category, but to go the distance on long tasks : planning multiple steps, calling tools reliably, recovering after an error, and delivering a usable result.


A model built for long-running agents

Most models behave well on an isolated prompt. A real agent has to do something else : make progress on an extended objective, with minimal supervision, navigating between tools and recovering from intermediate failures. That is the ground Grok 4.6 targets, both for software development copilots and for multi-step business workflows.

The model is also positioned beyond code, on structured engineering tasks : computer-aided design (CAD), procedural design, and complex technical analysis.


Highlighted benchmarks

Microsoft cites three evaluations in its announcement, all provided by the model publisher :

  • Terminal-Bench 3.0 : Measures the ability to carry out multi-step, terminal-based execution close to real engineering workflows.
  • 3DCodeBench : Evaluates agentic procedural 3D modeling through code, so the ability to reason about technical designs and generate them.
  • AA Briefcase : Artificial Analysis benchmark that evaluates agents on long knowledge-work projects, culminating in concrete deliverables such as spreadsheets, presentations, memos, financial models, or PDF files.

These results position Grok 4.6 on the workloads that produce a business deliverable, and not just an answer.


Deployment and regions

Grok 4.6 is one of the models sold by Azure, so it is not billed through the Azure Marketplace like partner models. Deployment is done from the Foundry portal catalog (ai.azure.com) by selecting the grok-4.6 model, or from the command line via Azure CLI.

The deployment characteristics are the following :

  • Deployment type : Global Standard only, in all supported regions
  • Context window : 200,000 tokens, input and output combined
  • Maximum output : 128,000 tokens, or the remaining context budget if that is smaller
  • Supported inputs : Text and image
  • API : Chat Completions and Responses
  • Authentication : Entra ID recommended, or an API key for the Foundry resource

The Cognitive Services Contributor role on the Foundry resource is required to deploy the model. Quotas depend on the subscription tier : 50,000 tokens per minute and 50 requests per minute on the Medium tier, 5,000,000 tokens per minute and 5,000 requests per minute on the High tier.

info

The direct SpaceXAI API advertises a 500,000 token context window for Grok 4.6, with pricing doubled beyond 200,000 tokens. In Foundry, the window is capped at 200,000 tokens : a request that exceeds this limit fails instead of switching to the long-context pricing.

Pricing

The rate in Foundry is aligned with the direct API, in Global Standard deployment :

  • Input : 2$ / 1M tokens
  • Output : 6$ / 1M tokens
  • Cache : 0.50$ / 1M tokens

At this price level, Grok 4.6 sits well below competing frontier models, which remains its main argument on token-heavy agentic workloads.


Why deploy it in Foundry ?

Access to the model is no longer the difficulty, operationalizing it is. Foundry brings a single platform to discover, evaluate, and deploy models, while applying the governance, security, and operational controls expected in the enterprise. Concretely, teams can evaluate Grok 4.6 against other frontier models on their own workloads, run workload-specific evaluations before deployment, expose it through managed endpoints, then integrate it into agentic applications without changing toolchain.


Conclusion

Grok 4.6 in Foundry is an additional option for teams building long-running agents, engineering copilots, or business automation, with an aggressive capability versus price ratio. The public preview status and the absence of an SLA that comes with it do require keeping it to evaluation phases, not yet to production workloads.


Sources

Microsoft - Techcommunity

Microsoft Learn - Deploy and use Grok models

SpaceXAI - Grok 4.6


Did you enjoy this post ? If you have any questions, comments or suggestions, please feel free to send me a message from the contact form.

Don’t forget to follow us and share this post.

Related Posts

Nearly 70% of Fortune 500 companies use Copilot

Nearly 70% of Fortune 500 companies use Copilot

Introduction At Microsoft Ignite 2024, Microsoft highlighted why nearly 70% of Fortune 500 companies now use Microsoft 365 Copilot. This mass adoption reflects a growing trend in the indu

Read More
How to disable self-service on Copilot licenses

How to disable self-service on Copilot licenses

Introduction Microsoft has activated a setting in the tenants (by default) to allow any user to purchase a Microsoft Copilot license through the *Microsoft 365 Copilot self-service pursha

Read More
Improved Teams video quality with Super Resolution

Improved Teams video quality with Super Resolution

Introduction Microsoft continues to innovate to provide users with the best possible virtual communication experience. One of the latest advancements is the introduction of *Super Resolutio

Read More
Le Chat by Mistral AI, your personal AI assistant

Le Chat by Mistral AI, your personal AI assistant

Introduction I told you last December about the French AI, Mistral AI, the most popular model in Europe in which Microsoft invested 15 million euros in the startup. The mobile app has jus

Read More
New Yealink MeetingBoard 65 and 85 for Teams rooms

New Yealink MeetingBoard 65 and 85 for Teams rooms

Introduction The new Yealink MeetingBoard 65 and 85 are an innovative and comprehensive solution designed to transform meeting rooms into intelligent collaboration spaces. These all-in-on

Read More
Maximize the use of the Copilot prompt gallery

Maximize the use of the Copilot prompt gallery

Introduction Microsoft 365 Copilot continues to revolutionize the way organizations work by integrating advanced artificial intelligence capabilities into everyday tools. One of the key f

Read More
How to get started with Copilot in Excel

How to get started with Copilot in Excel

Introduction Microsoft 365 Copilot is a major innovation that integrates artificial intelligence directly into the applications you use every day, like Excel. Copilot helps you automate t

Read More
Microsoft Purview for Azure Data Lake and Blob Storage

Microsoft Purview for Azure Data Lake and Blob Storage

Introduction Microsoft announced that Microsoft Purview protection policies for Azure Data Lake and Blob Storage are now available in all regions. This advancement allows organization

Read More
Facilitator, new AI agent for taking notes in meetings

Facilitator, new AI agent for taking notes in meetings

Introduction Microsoft recently announced a new feature for Teams Rooms: Facilitator ; an AI agent that takes notes during Teams meetings. This feature is currently in pre-public release

Read More
Enterprise Connect 2025 : Yealink SkySound CM50 Dante kit

Enterprise Connect 2025 : Yealink SkySound CM50 Dante kit

Introduction Enterprise Connect is an annual conference that brings together communications technology professionals, innovators, and others. This event showcases technological advances i

Read More
Mistral OCR, new benchmark in character recognition

Mistral OCR, new benchmark in character recognition

Introduction In March 2025, Mistral AI announced the launch of Mistral OCR, an optical character recognition (OCR) API that sets a new standard in document understanding. This advance

Read More
Introducing the Logitech Rally Board 65

Introducing the Logitech Rally Board 65

Introduction The Logitech Rally Board 65 is an all-in-one video conferencing solution designed to simplify meetings and collaboration in business environments. With its 65-inch touchscree

Read More
Mistral Code, the European AI development assistant

Mistral Code, the European AI development assistant

Introduction French startup Mistral AI, already recognized for its open source language models, has just unveiled Mistral Code, an intelligent development assistant designed for businesse

Read More
New Yealink MeetingBar A50 for Teams Rooms

New Yealink MeetingBar A50 for Teams Rooms

Introduction In an increasingly hybrid work world, businesses are looking for video conferencing solutions that are powerful, easy to deploy, and seamlessly integrated into their *Microsoft

Read More
Mercedes-Benz, your car becomes a rolling office

Mercedes-Benz, your car becomes a rolling office

Introduction In an automotive market increasingly focused on smart and connected mobility, Mercedes-Benz is taking a giant leap forward. With the new generation of the CLA model, the Ge

Read More
Anthropic unveils Claude Opus 4.1, faster and more reliable

Anthropic unveils Claude Opus 4.1, faster and more reliable

Introduction Anthropic, a leading player in artificial intelligence, has announced the release of Claude Opus 4.1, a significant update to its flagship model (Claude Opus 4). Designed

Read More
OpenAI unveils GPT-5, its latest smarter model

OpenAI unveils GPT-5, its latest smarter model

Introduction OpenAI has taken another step forward in the evolution of artificial intelligence with the launch of GPT-5, its most powerful language model to date. Designed to be smarter

Read More
What's new for Copilot in August 2025

What's new for Copilot in August 2025

Introduction Microsoft releases a monthly update to Microsoft 365 Copilot to keep admins and users up-to-date on productivity-enhancing features in Microsoft 365. The August 2025 release

Read More
Anthropic unveils Claude Sonnet 4.5, more advanced

Anthropic unveils Claude Sonnet 4.5, more advanced

Introduction Anthropic, a leading player in artificial intelligence, has announced the release of Claude Sonnet 4.5, touted as the world's best coding model and a significant leap for

Read More
How to enable DSPM for AI with Purview

How to enable DSPM for AI with Purview

Introduction With the rise of generative AI models, the phenomenon of Shadow AI (the use of artificial intelligence tools and services not approved or controlled by organizations) is incr

Read More
How to add a disclaimer in Copilot

How to add a disclaimer in Copilot

Introduction Microsoft has enabled a setting in tenants that allows administrators to display the Microsoft 365 Copilot disclaimer in bold, and to attach a shortcut pointing to a usage po

Read More
Extend Zero Trust to AI agent identities in Entra ID

Extend Zero Trust to AI agent identities in Entra ID

Introduction AI agents are becoming increasingly widespread in businesses (incident summaries, log analysis, flow execution, etc.), and it is crucial that their access is continuously evalu

Read More
Mistral Voxtral Transcribe2, real-time transcription

Mistral Voxtral Transcribe2, real-time transcription

Introduction Mistral AI has just unveiled Voxtral Transcribe 2, its second generation of speech transcription models with cutting-edge transcription quality, ultra-low latency and advan

Read More
How to enable DLP for AI websites with Purview

How to enable DLP for AI websites with Purview

Introduction Last week, I showed you how to enable DLP to prevent printing of financial data using Microsoft Purview, in order to prevent accidental or malicious data leaks (*Data Loss

Read More
Anthropic unveils Claude Opus 4.6, a benchmark for finance

Anthropic unveils Claude Opus 4.6, a benchmark for finance

Introduction Artificial intelligence is rapidly growing in the finance industry, but one reality remains : real-world financial analyses are rarely clean, linear, or perfectly defined. They

Read More
How to enable Claude AI as a model in Copilot

How to enable Claude AI as a model in Copilot

Introduction Since its launch, Microsoft 365 Copilot has established itself as a cornerstone of enhanced enterprise productivity, leveraging advanced AI models to reason, analyze, and aut

Read More
OpenAI unveils GPT-5.4, the new generation of models

OpenAI unveils GPT-5.4, the new generation of models

Introduction OpenAI has just announced GPT-5.4, a new evolution of its GPT model family. Designed for professional uses and complex tasks, this model introduces several major improvemen

Read More
Introducing Microsoft 365 E7, the Frontier Suite

Introducing Microsoft 365 E7, the Frontier Suite

Introduction Microsoft has announced the availability of the Microsoft 365 E7 license, a new offer called Frontier Suite, designed for the era of AI-driven work and agents. This announc

Read More
OpenAI unveils GPT-5.5, designed for agentic work

OpenAI unveils GPT-5.5, designed for agentic work

Introduction OpenAI has just announced GPT-5.5, barely seven weeks after GPT-5.4. The message is clear, GPT-5.5 is not an incremental update but *"a new class of intelligence for real

Read More
Mistral Small 4, one model for reasoning, visioning, and coding

Mistral Small 4, one model for reasoning, visioning, and coding

Introduction Managing multiple specialized models within a single AI pipeline adds deployment complexity and multiplies infrastructure costs. Mistral AI announced on March 16, 2026 the la

Read More
Voice-native agents in Foundry in Public Preview

Voice-native agents in Foundry in Public Preview

Introduction Microsoft announced on March 16, 2026 the Public Preview of Voice Native Agents in Microsoft Azure AI Foundry, a native combination of the Voice Live API and the *Found

Read More
Anthropic unveils Claude Opus 4.7, with a new tokenizer

Anthropic unveils Claude Opus 4.7, with a new tokenizer

Introduction Anthropic announced on April 16, 2026 the General Availability of Claude Opus 4.7, the direct successor to [Claude Opus 4.6](https://maxime.hiez.ca/en/blog/2026-02-13-ai-

Read More
GPT-5.5 Instant now available in Copilot

GPT-5.5 Instant now available in Copilot

Introduction Microsoft 365 Copilot is accelerating the renewal cadence of its underlying models ; this is the fifth iteration of the GPT-5.x series deployed in less than a year. On 7 May 202

Read More
Consecutive bilingual interpretation in Teams

Consecutive bilingual interpretation in Teams

Introduction The Interpreter agent in Microsoft Teams previously offered only a simultaneous translation mode, where the synthesized voice overlapped with the speaker's voice in real ti

Read More
Anthropic releases Claude Opus 4.8

Anthropic releases Claude Opus 4.8

Introduction Anthropic made Claude Opus 4.8 generally available on May 28, 2026, as the direct successor to Claude Opus 4.7. Pricing remains unchanged at 5$ / 1M input tokens and 25$

Read More
Claude Opus 4.8 available in Microsoft Foundry

Claude Opus 4.8 available in Microsoft Foundry

Introduction Microsoft made Claude Opus 4.8 available in Microsoft Foundry in preview, a few days after its general release by Anthropic on May 28, 2026. The model is accessible via

Read More
Anthropic launches Claude Fable 5 and Claude Mythos 5

Anthropic launches Claude Fable 5 and Claude Mythos 5

Introduction Anthropic announces the launch of Claude Fable 5 and Claude Mythos 5, two models built on the same foundation, with capabilities surpassing those of any previously releas

Read More
OpenAI unveils GPT-5.6 Sol, Terra and Luna in preview

OpenAI unveils GPT-5.6 Sol, Terra and Luna in preview

Introduction OpenAI announced on June 26, 2026 a preview of three new models : GPT-5.6 Sol, GPT-5.6 Terra and GPT-5.6 Luna. Access is limited to approximately twenty trusted partner

Read More
Anthropic unveils Claude Sonnet 5, outperforming Opus 4.8

Anthropic unveils Claude Sonnet 5, outperforming Opus 4.8

Introduction Anthropic announced on June 30, 2026, the availability of Claude Sonnet 5, now the default model on the Free and Pro plans of Claude.ai. The model is designed for com

Read More
Anthropic releases Claude Opus 5

Anthropic releases Claude Opus 5

Introduction Anthropic made Claude Opus 5 generally available on July 24, 2026, as the successor to Claude Opus 4.8. Pricing remains unchanged at 5$ / 1M input tokens and 25$ / 1M out

Read More
Claude Opus 5 available in Microsoft Foundry

Claude Opus 5 available in Microsoft Foundry

Introduction Claude Opus 5 has been available in Microsoft Foundry since July 24, 2026, simultaneously with its general release by Anthropic. For the first time, a Claude model arri

Read More
Teams Phone Agent and Copilot Studio voice agents in Preview

Teams Phone Agent and Copilot Studio voice agents in Preview

Introduction Microsoft Teams Auto Attendants have always relied on a fixed menu, where the caller listens to a list of options before pressing a key. Teams Phone Agent replaces that p

Read More
GPT-5.6 now available in Copilot

GPT-5.6 now available in Copilot

Introduction On July 9, 2026, GPT-5.6 became the preferred model of Microsoft 365 Copilot, on the very day of its general availability at OpenAI. The preview of the range, restricted

Read More
Shieldstral, the multimodal safety classifier from Mistral AI

Shieldstral, the multimodal safety classifier from Mistral AI

Introduction Mistral AI announced on August 4, 2026 Shieldstral, a multimodal safety classifier meant to moderate the inputs and the outputs of a language model. The model sits before a

Read More
Mistral Agentic Search, the iterative document search

Mistral Agentic Search, the iterative document search

Introduction Mistral AI announced on August 20, 2026 Agentic Search, a document search layer meant to replace the classic RAG in the cases where it fails. The principle changes : inst

Read More
OpenAI unveils Ultrafast mode for GPT-5.6 Sol in preview

OpenAI unveils Ultrafast mode for GPT-5.6 Sol in preview

Introduction OpenAI announced on August 13, 2026 Ultrafast, a new API service tier that runs GPT-5.6 Sol up to 14 times faster than standard processing. The model is the same, with

Read More
Anthropic launches Claude Fable 5.1 and Claude Mythos 5.1

Anthropic launches Claude Fable 5.1 and Claude Mythos 5.1

Introduction Anthropic announces Claude Fable 5.1 and Claude Mythos 5.1, two variants of a single model succeeding Claude Fable 5 and Claude Mythos 5. Beyond the capabilities, the

Read More
Claude Fable 5.1 available in Microsoft Foundry

Claude Fable 5.1 available in Microsoft Foundry

Introduction Microsoft expands its Anthropic catalog in Microsoft Foundry with Claude Fable 5.1 and Claude Mythos 5.1, the same day as their launch by Anthropic. As with *Claude O

Read More
OpenAI unveils GPT-6 Astra

OpenAI unveils GPT-6 Astra

Introduction OpenAI announces GPT-6 Astra, presented as its most intelligent and best-aligned model to date. Astra saturates several reference benchmarks : 98% on FrontierMath Tier 4,

Read More
GPT-6 Astra available in Copilot

GPT-6 Astra available in Copilot

Introduction GPT-6 Astra joins the frontier models available in Copilot Cowork and Copilot Studio, five days after its launch by OpenAI. The rollout began on September 4, 2026. Che

Read More