top of page
Search

All Copilot models available in 2025: full list for web, app, Microsoft 365, GitHub, Studio, and API integration

ree

Copilot now includes GPT-5 and custom routing across all Microsoft platforms, from Office to developer APIs.

As of August 2025, Microsoft’s Copilot ecosystem spans a variety of products powered by orchestrated AI models. From consumer chat experiences in Copilot (web and app) to enterprise productivity tools in Microsoft 365 and developer workflows via GitHub and Copilot Studio, the platform utilizes a combination of OpenAI models, Anthropic, and Google Gemini, routed dynamically or selectable directly depending on the service tier.


Here we outline all the models currently available across the different Copilot environments, including model versions, selection methods, and platform-specific configurations.



Copilot web, desktop, and mobile apps use GPT-5 via Smart Mode.

On copilot.microsoft.com and the Copilot apps (Windows, Mac, iOS, Android), model selection is automated through Microsoft’s Smart Mode. This mode dynamically routes each prompt to the most appropriate LLM, based on task type and user behavior.


● GPT-5 (Smart Mode default)

Currently the primary model behind all Copilot user-facing experiences. Smart Mode automatically invokes GPT-5 and, when needed, enhanced variants like “GPT-5 Thinking” for reasoning-heavy inputs. Users do not need to manually select this model—it is triggered by prompt complexity.


● GPT-4o (fallback or hybrid use)

Although now secondary, GPT-4o still serves as a fallback model or backend for some faster, latency-sensitive queries. It also participates in multi-model orchestration managed by Microsoft’s Prometheus/Orchestrator engine, which combines LLM reasoning with real-time web grounding.



Microsoft 365 Copilot uses multi-LLM routing with orchestration.

In the enterprise-focused Microsoft 365 Copilot (Word, Excel, Outlook, PowerPoint, Teams), AI requests are routed through a sophisticated LLM orchestration system. This engine integrates:

  • GPT-5 for structured reasoning, writing, and summarization

  • Prometheus Layer for Bing-based grounding

  • Microsoft Graph for personalized enterprise data access

  • Model switching logic depending on product and task (e.g., charts in Excel vs. emails in Outlook)

Users cannot select models directly; all routing is invisible and contextual. However, Microsoft has confirmed that GPT-5 is now in active use across the 365 Copilot suite as of mid-2025.



Copilot Studio allows manual model selection and Azure OpenAI linkage.

Copilot Studio, Microsoft's low-code platform for building custom copilots and AI agents, offers users direct control over the LLMs powering generative responses.


● GPT-4o (default)

The standard model used for generative answers in Studio-created copilots. It provides stable, well-balanced completions suitable for business logic, workflows, and Q&A.


● GPT-4.5 (Preview)

A preview model that builders can opt into. It offers improved logic and summarization capabilities but is not recommended for production deployments due to potential versioning changes.


● Azure OpenAI Integration

Studio agents can be connected to a dedicated Azure OpenAI resource, allowing access to a wide range of GPT models, including:

  • GPT-5, GPT-5-mini, GPT-5-chat

  • GPT-4.1, GPT-4o, o3, o3-pro, o4-mini

  • Multimodal APIs for audio (gpt-4o-realtime-preview), TTS, and image generation (dall-e-3, sora)


This configuration supports granular control over cost, temperature, model versioning, and security compliance.



GitHub Copilot offers multi-model options with OpenAI, Anthropic, and Google Gemini.

In GitHub Copilot (IDE integrations, GitHub.com, mobile), model selection is partially automated but increasingly customizable under GitHub Copilot for Business and Enterprise tiers.

Available models include:


● OpenAI models

  • GPT-5 (Preview)

  • GPT-4.1 (GA)

  • GPT-4o, o3, o3-pro, o4-mini


● Anthropic models

  • Claude Opus 4.1 (Preview)

  • Claude Opus 4.0, Sonnet 4.0, Sonnet 3.7, Sonnet 3.5


● Google models

  • Gemini 2.5 Pro (Preview)

  • Gemini 2.0 Flash (GA)


GitHub Copilot dynamically selects models for completions, documentation, and code chat, but admins may control model routing in enterprise environments for performance or compliance reasons.



Azure OpenAI provides full model access for developers and enterprise AI workflows.

For users building outside of the Copilot interface, Azure OpenAI offers full API-level access to Microsoft’s supported LLMs. The latest model lineup (as of August 2025) includes:

Model Name

Model ID / Alias

Status

GPT-5

gpt-5, gpt-5-mini, gpt-5-chat

GA / Preview

GPT-4.1

gpt-4.1, gpt-4.1-mini, gpt-4.1-nano

GA

GPT-4o

gpt-4o, gpt-4o-mini, gpt-4o-realtime-preview

GA

o-series

o3, o3-pro, o3-mini, o4-mini

GA

Audio & TTS

gpt-4o-mini-tts, gpt-4o-preview-native-audio-dialog

Preview

Image / Video

dall-e-3, gpt-image-1, sora

Preview

Azure users can deploy these models via REST API, Azure SDKs, or embedded Copilot Studio agents. This route is preferred for production AI services requiring control over quota, region, latency, and regulatory scope.



Choosing the right Copilot model depends on your product, plan, and control level.

With routing and orchestration deeply embedded into Microsoft’s ecosystem, most users do not need to manually choose models. However, here is a breakdown of when each model family becomes relevant:

  • For general productivity and chat → Copilot (web/app) with Smart Mode (uses GPT-5)

  • For Office documents, Teams, Excel, email → Microsoft 365 Copilot (uses orchestrated GPT-5)

  • For custom AI agents and workflows → Copilot Studio with GPT-4o or Azure GPT-5

  • For coding and developer tools → GitHub Copilot with multi-LLM routing

  • For enterprise AI deployments → Azure OpenAI with full GPT model suite and APIs


Microsoft’s Copilot infrastructure has quietly become one of the most sophisticated AI routing systems in production, blending OpenAI’s LLMs with enterprise logic, latency tuning, and seamless product integration. A technical comparison chart of all Copilot-accessible models, with latency, quota, and pricing data, can be provided upon request.



____________

FOLLOW US FOR MORE.


DATA STUDIOS


bottom of page