At its Build developer conference, Microsoft announced an expansion of its model family and agentic-AI infrastructure, and made the unified intelligence layer Microsoft IQ generally available, according to SiliconANGLE. SiliconANGLE reports that Work IQ APIs will be available on June 16 to give agents direct access to Microsoft 365 signals, and that a Fabric-hosted semantic data foundation will act as an ontology for structured business data. The company also introduced a web-grounding capability it says is model-agnostic and Model Context Protocol native, claiming it returns relevant information blocks about 2.5 times faster than alternatives, per SiliconANGLE. Reuters covered related announcements including prototype hardware under "Project Solara" for running AI agents on-device and a Satya Nadella quote on the new platform rules.
What happened
Microsoft announced an expansion of its AI model families and agentic-AI infrastructure at the Build developer conference in San Francisco, per SiliconANGLE and Reuters. SiliconANGLE reports that Microsoft IQ is now generally available. SiliconANGLE says the offering includes Work IQ APIs, which the article reports will become available on June 16 to provide agents direct access to signals inside Microsoft 365. SiliconANGLE also describes a Fabric-hosted semantic data foundation that it reports will serve as an "ontology" for structured business data and tie together unstructured sources such as wikis, policies and the live web.
Technical details
SiliconANGLE reports that the Microsoft Superintelligence Team released a family of seven new in-house models, led by the reasoning model MAI-Thinking-1. According to SiliconANGLE, MAI-Thinking-1 is a 35 billion-parameter model with a 128,000-token context window. SiliconANGLE attributes to the company the claim that independent raters preferred MAI-Thinking-1 to Anthropic PBC's Sonnet 4.6 in a blind test, and that it matched Opus 4.6 on coding performance in SWE Bench Pro. SiliconANGLE also reports Microsoft introduced a model-agnostic web-grounding capability, described as Model Context Protocol native, which the article says the company claims returns relevant information blocks about 2.5 times faster than the next-best alternative.
Related product and hardware announcements
Reuters reports Microsoft revealed prototype hardware under the name Project Solara, a family of devices designed to host AI agents rather than run traditional operating systems and apps. Reuters says the prototypes include devices roughly the size of a smart speaker or badge, built on chips from Qualcomm and MediaTek, and that Microsoft showcased use cases such as documenting a medical visit. Reuters quotes CEO Satya Nadella: "It's a new platform, but perhaps more importantly, it's a set of new platform rules that don't, in some sense, hem in what you can imagine."
Industry context
Editorial analysis: Microsoft's announcements combine three developer-facing threads: larger-context models, agentic infrastructure that connects models to enterprise signals, and edge or near-edge device prototypes that run agents. Firms building agent platforms typically prioritize long context windows, retrieval and grounding layers, and APIs that expose enterprise signals. Observers following the sector have seen similar vendor moves where integrated semantic layers and grounding are positioned as the primary defense against hallucinations.
For practitioners
Editorial analysis: The practical implications for ML engineers and platform teams are twofold. First, supporting 128,000-token contexts and agentic workflows increases demands on retrieval, caching, and prompt-construction tooling. Second, the introduction of Fabric-hosted semantic layers and Work IQ APIs highlights a continuing trend toward standardized enterprise connectors and ontologies for grounding. Organizations evaluating agentic designs should watch model-context protocols and how vendors expose enterprise signals via APIs.
What to watch
- •Adoption timing and availability: whether Work IQ APIs roll out as reported on June 16, and the initial SDK and SLA terms. - •Independent benchmark details: third-party evaluations that corroborate or dispute the MAI-Thinking-1 comparisons to Sonnet 4.6 and Opus 4.6.
- •Project Solara form factors and developer tooling required to run agentic workloads on-device, including chip partners and cloud integration.
Scoring Rationale #
Major vendor announcements affect platform and tool choices for practitioners: new reasoning models with very long context windows and agentic APIs change retrieval, prompt engineering, and orchestration design. The inclusion of prototype hardware increases strategic relevance for edge and hybrid deployments.
Practice with real Ride-Hailing data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card