Microsoft has announced the availability of its latest AI models, o3 and o4-mini, on the Azure OpenAI Service. These models mark a substantial improvement in AI reasoning, offering enhanced quality, safety, and performance compared to their predecessors. The new models are available through Azure AI Foundry and GitHub, enabling developers to create more sophisticated AI-powered experiences.
Key Enhancements
Both o3 and o4-mini models bring significant upgrades in quality and safety while maintaining support for existing features from previous models. The key features include:
- Multiple API Support: Both models are compatible with Responses API and Chat Completions API. The Responses API allows for seamless integration with multiple tools and provides enhanced transparency through a reasoning summary in the model output.
- Reasoning Summary: The Responses API now includes a reasoning summary in its output, offering deeper insights into the model’s thought process and improving the effectiveness of actions and tools that utilize these insights.
- Multimodality: o3 and o4-mini expand their reasoning capabilities to process and analyze visual data. o3 features enhanced vision analysis, while o4-mini introduces new vision support, allowing for more comprehensive text outputs based on visual information.
- Full Tools Support: Both models support parallel tool calling and are the first reasoning models with full tools support, similar to mainline models. This capability enables customers to develop advanced agentic solutions.
Safety Improvements
The o-series reasoning models utilize a training strategy called deliberative alignment, which teaches the models to reason explicitly about safety specifications before providing answers. Both o3 and o4-mini represent the next level of safety improvements within the o-series, ensuring that users can leverage their capabilities while maintaining high safety standards.
Additional AI Innovations
Alongside the o3 and o4-mini models, Azure OpenAI Service has introduced three new audio models available for deployment in East US2 on Azure AI Foundry. These include GPT-4o-Transcribe, GPT-4o-Mini-Transcribe for speech-to-text applications, and GPT-4o-Mini-TTS, a customizable text-to-speech model. These models offer superior performance and customization options for audio processing tasks.
Embracing the Future of AI Reasoning
The introduction of o3 and o4-mini models signifies a new era in AI reasoning, where AI becomes a more integral partner in innovation. These models are not mere upgrades but gateways to new possibilities in AI applications. They enable users to tackle complex problems, create seamless workflows, and explore new frontiers in AI. With Azure OpenAI Service, users can now elevate their AI journey and push the boundaries of what AI can achieve. Developers can sign up to use o3 and o4-mini models in Azure AI Foundry today and start creating more advanced AI-powered solutions.
