Microsoft

MAI-Cyber-1-Flash

Microsoft

MAI-Cyber-1-Flash is a compact security model in Microsoft’s MAI-Thinking-1 lineage, offered to verified defenders through the MDASH vulnerability-analysis and remediation harness. It handles routine code-security tasks while MDASH routes difficult cases to larger models. The documented August release is a controlled enterprise offering rather than a general downloadable or standalone public API model. Reported CyberGym results evaluate a combined system using this model and GPT-5.4, so those numbers are not assigned to this model alone.

MAI-Thinking-1

Microsoft

MAI-Thinking-1 is Microsoft AI’s medium-inference-footprint reasoning model, first announced at Build in June and made available in public preview on Microsoft Foundry on August 12, 2026. It uses a sparse mixture-of-experts architecture and was trained in-house without third-party model distillation. The model targets mathematical reasoning, software engineering, and complex enterprise questions. Clean-data and benchmark statements are provider claims rather than guarantees of domain correctness. June announcement-only exposure is not used as its public release date.

MAI-Code-1.1-Flash

Microsoft

MAI-Code-1.1-Flash is a substantive update to Microsoft’s small coding model deployed in GitHub Copilot. It improves terminal-agent and .NET tasks and was trained from production feedback with reinforcement-learning environments tied to developer workflows. Microsoft reports improvements in token efficiency, streaming speed, and code retention relative to the June model. These relative production measurements are not interchangeable with standalone benchmark accuracy. The model is accessed through Copilot and does not have publicly downloadable weights.

MAI-Code-1-Flash

Microsoft

MAI-Code-1-Flash is Microsoft AI’s small active-parameter coding model, deployed in GitHub Copilot for everyday developer work. It was trained together with the production Copilot harness to handle repository questions, refactoring, multi-turn code changes, and adaptive reasoning. Microsoft reports five billion active parameters but has not disclosed the full architecture or total size. The model remains proprietary; access in a developer product is distinct from downloadable weights or a generally available independent API.

Phi-4-reasoning-vision-15B

Microsoft

Phi-4-reasoning-vision-15B is an MIT-licensed open-weight model in Microsoft’s Phi family. It adds visual perception to the Phi-4-reasoning backbone, supporting visual math, chart interpretation, screenshots, and image-conditioned reasoning. Weights and sample inference code are available on Hugging Face, with deployment through supported local runtimes or hosted platforms. This checkpoint is a static trained model; accuracy and language coverage vary by task. Small model efficiency should not be taken as equal capability to much larger generalist systems.

MAI-1-preview

Microsoft

MAI-1-preview is Microsoft AI’s first foundation language model trained end-to-end in-house, initially offered for public testing on LMArena. It is an instruction-oriented mixture-of-experts model for everyday consumer questions and helpful responses. Microsoft also invited trusted testers to apply for API access and announced gradual integration into selected Copilot text use cases. Public testing qualifies as preview availability, but neither unrestricted API access nor downloadable weights were released with it.

Phi-4-mini-flash-reasoning

Microsoft

Phi-4-mini-flash-reasoning is an MIT-licensed open-weight model in Microsoft’s Phi family. A new SambaY architecture uses state-space modeling and gated memory sharing to improve decoding efficiency for compact mathematical reasoning. Weights and sample inference code are available on Hugging Face, with deployment through supported local runtimes or hosted platforms. This checkpoint is a static trained model; accuracy and language coverage vary by task. Small model efficiency should not be taken as equal capability to much larger generalist systems.

Phi-4-mini-reasoning

Microsoft

Phi-4-mini-reasoning is an MIT-licensed open-weight model in Microsoft’s Phi family. The compact reasoning model specializes in mathematical reasoning through curated and synthetic training examples. Weights and sample inference code are available on Hugging Face, with deployment through supported local runtimes or hosted platforms. This checkpoint is a static trained model; accuracy and language coverage vary by task. Small model efficiency should not be taken as equal capability to much larger generalist systems.

Phi-4-reasoning

Microsoft

Phi-4-reasoning is an MIT-licensed open-weight model in Microsoft’s Phi family. It fine-tunes Phi-4 with curated reasoning demonstrations. Reasoning-plus adds outcome-based reinforcement learning and longer reasoning traces. Weights and sample inference code are available on Hugging Face, with deployment through supported local runtimes or hosted platforms. This checkpoint is a static trained model; accuracy and language coverage vary by task. Small model efficiency should not be taken as equal capability to much larger generalist systems.

Phi-4-mini

Microsoft

Phi-4-mini is an MIT-licensed open-weight model in Microsoft’s Phi family. The Mini release adds an expanded vocabulary, grouped-query attention, and improved instruction tuning for compact multilingual and tool-calling applications. Weights and sample inference code are available on Hugging Face, with deployment through supported local runtimes or hosted platforms. This checkpoint is a static trained model; accuracy and language coverage vary by task. Small model efficiency should not be taken as equal capability to much larger generalist systems.

Phi-4-multimodal

Microsoft

Phi-4-multimodal is an MIT-licensed open-weight model in Microsoft’s Phi family. It combines the Phi-4-mini language backbone with vision and speech encoders and adapters, producing text from mixed inputs. Weights and sample inference code are available on Hugging Face, with deployment through supported local runtimes or hosted platforms. This checkpoint is a static trained model; accuracy and language coverage vary by task. Small model efficiency should not be taken as equal capability to much larger generalist systems.

Phi-4

Microsoft

Phi-4 is an MIT-licensed open-weight model in Microsoft’s Phi family. It emphasizes high-quality synthetic and curated data, with particular strength in STEM and mathematical problem solving. Weights and sample inference code are available on Hugging Face, with deployment through supported local runtimes or hosted platforms. This checkpoint is a static trained model; accuracy and language coverage vary by task. Small model efficiency should not be taken as equal capability to much larger generalist systems.

Phi-3

Microsoft

14B-parameter follow-up to Phi-1/2 focusing on compact "small models" with high-quality training. Released in sizes 3B, 7B, 14B. Emphasizes efficient training and inference.

Phi-2

Microsoft

2.7B-parameter model focused on efficiency and quality by training on high-quality (including synthetic) data. Achieved surprisingly strong results for its size, dubbed a "textbook-quality" model.

Megatron-Turing NLG 530B

Microsoft

A 530-billion-parameter Transformer language model, one of the largest dense LLMs of 2021. Collaboration between NVIDIA and Microsoft, used for research on scaling.