Recap of what's new in the OpenAI API

Happy New Year! We ended last year with DevDays in SF, London, Singapore, and a holiday edition online. You can watch the videos on YouTube now. We’re particularly excited for what’s ahead in 2025—we’ve a lot in store to help you build with OpenAI.?

Here’s a quick recap of everything we launched recently.

Product Updates

OpenAI o1 in the API

We’re rolling out OpenAI o1 to the API, with support for function calling, Structured Outputs, vision, and developer messages. We’ve also added a new reasoning_effort parameter, which lets you control how long the model thinks before answering. And compared to o1-preview, it’s lower latency—o1 uses 60% fewer reasoning tokens on average. o1 is rolling out to developers on usage tier 5 and we’re working on expanding to more tiers soon.

Realtime API improvements

The Realtime API now points to our latest model snapshot gpt-4o-realtime-preview-2024-12-17 by default, which drops the audio token price by 60% and cached audio input by 87.5%. We’ve also improved voice quality and made inputs more reliable. We’ve also brought GPT-4o mini to the Realtime API, which is up to 10x cheaper than previous prices. And on top of that, the Realtime API now supports WebRTC, which can make your integration just a handful of lines of code. Learn more in our updated guide.

Preference fine-tuning

We introduced preference fine-tuning, which lets you customize models based on comparing responses, rather than fixed targets. This method is called Direct Preference Optimization, which teaches the model to distinguish between your preferred and non-preferred outputs. It’s especially effective for subjective tasks where tone, style, and creativity matter. Preference fine-tuning is available for GPT-4o, and we’re working on support for more models this year. Get started with our docs.

Usage API

You can now track API usage and costs programmatically with our new Usage API. Monitor token use by minute/hour/day, and filter usage by API key, project ID, user ID, or model. The Costs endpoint also lets you keep tabs on your daily spend. Learn how to track usage in our API reference.

New Developer Resources

Building with the Realtime API

• Realtime API Solar System demo
• Realtime API agents demo
• Realtime API console with WebRTC
• Realtime API with Twilio
• Realtime API embedded SDK for microcontrollers

Sample apps

• Structured Outputs sample apps

SDKs and docs

• Improved function calling guide
• Go SDK beta
• Java SDK beta

Community

This year, we want to meet as many of you in person as possible. If you’re interested in coming to a developer meetup near you, let us know in this form. And as always, if you have any questions, suggestions, or feedback, please join us in the developer forum.