AI BriefingOpenAI Releases o1 and New Developer Tools
Key point
OpenAI has unveiled new tools for developers, including API support for the reasoning model **o1** and improvements to the **Realtime API**.
Details
OpenAI is now offering the reasoning model o1 via API, along with updates to a variety of features and tools for developers. This update includes improvements to the Realtime API, a new Preference Fine-Tuning method, and beta versions of the Go and Java SDKs.
The o1 API supports the following key features optimized for production environments:
- Function calling: seamless connection with external data and APIs
- Structured Outputs: reliable responses matching user-defined JSON Schemas
- Developer messages: specifying the model's tone, style, and other behavioral guidelines
- Vision capabilities: support for a variety of applications through image-based reasoning
- Reduced latency: an average of 60% fewer reasoning tokens compared to o1-preview
- reasoning_effort parameter: allows control over the model's thinking time
The newly released o1-2024-12-17 snapshot improves both performance and cost efficiency over the previous version. In particular, it achieved dramatic performance gains on major benchmarks such as SWE-bench Verified and AIME 2024, setting a new SOTA (State-of-the-art).
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.