Microsoft Azure OpenAI Service Launches Advanced AI Reasoning Models o3 and o4-mini

Microsoft has announced the availability of o3 and o4-mini, the latest iterations in its o-series of AI reasoning models, now accessible via the Microsoft Azure OpenAI Service and GitHub. These models mark a significant advancement in AI reasoning, delivering improved quality, safety, and performance.

Key Features and Enhancements

• Multiple APIs Support: Both models are compatible with Responses API and Chat Completions API, enabling seamless integration with various tools.
• Reasoning Summary: The Responses API now includes a reasoning summary in the output, enhancing transparency and explainability.
• Multimodality: Enhanced vision analysis capabilities in o3 and new vision support in o4-mini allow for processing visual data and generating comprehensive text outputs.
• Full Tools Support: Both models support parallel tool calling, enabling the development of advanced agentic solutions.

Safety Innovations

The o-series models employ deliberative alignment, a training strategy that ensures models reason explicitly about safety specifications before responding. o3 and o4-mini incorporate the latest safety improvements, reinforcing their reliability for enterprise use.

New Audio Models

Azure OpenAI Service has also introduced three new audio models:

• GPT-4o-Transcribe and GPT-4o-Mini-Transcribe: Advanced speech-to-text models outperforming previous benchmarks.
• GPT-4o-Mini-TTS: A customizable text-to-speech model allowing detailed control over speech characteristics. Learn more on the Tech Community blog.

A New Era in AI Reasoning

With o3 and o4-mini, Microsoft is ushering in a new era of AI reasoning, enabling businesses to tackle complex problems and innovate with confidence. Sign up to use o3 and o4-mini in Azure AI Foundry today.