OpenAI Unveils Advanced Reasoning Models: o3 and o4-mini
Today marks a significant advance in artificial intelligence as OpenAI introduced its cutting-edge reasoning models, o3 and o4-mini. These models demonstrate a remarkable improvement over their earlier iterations, o1 and o3-mini, setting a new benchmark for AI performance.
Availability and Pricing
Microsoft has announced the release of these new models via the Azure OpenAI Service within the Azure AI Foundry. The pricing structure is as follows:
- o3 Model: $10 per million input tokens; $40 per million output tokens.
- o4-mini Model: $1.10 per million input tokens; $4.40 per million output tokens.
More big updates to Foundry today: o3 and o4-mini from OpenAI are both simul-shipping, delivering a leap forward in AI reasoning.https://t.co/gT3rtAhv3i
— Satya Nadella (@satyanadella) April 16, 2025
Enhanced Features: Vision and Tool Support
The implementations of these models come with exciting new features, including enhanced vision capabilities. For the first time, users can submit image inputs, which are supported in both the Responses API and Chat Completions API. Additionally, both models now support comprehensive tools and facilitate parallel tool calling, enhancing their functionality.
New Audio Models and GitHub Integration
In conjunction with the introduction of o3 and o4-mini, Azure OpenAI Service has rolled out new audio models in the East US2 region at Azure AI Foundry, including:
- GPT-4o-Transcribe
- GPT-4o-Mini-Transcribe
- GPT-4o-Mini-TTS
Moreover, Microsoft’s GitHub has confirmed that both reasoning models are now publicly available in GitHub Copilot and GitHub Models. The o4-mini model will be accessible across all paid GitHub Copilot plans, while the o3 model will be rolling out for Enterprise and Pro+ plans.
Implementation and Access for Developers
Once fully deployed, users will have the ability to select these new models via the model picker within Visual Studio Code or the GitHub Copilot Chat on GitHub. Additionally, for organizations using Copilot Enterprise, administrators will need to enable access through updated policies in Copilot settings.
These innovative models are also integrated within GitHub Models, empowering developers to explore, build, and deploy AI-driven features seamlessly alongside other Microsoft and third-party models.
Leave a Reply ▼