We’ve learned so much firsthand from developers building on @OpenAI
’s frontier models who also turn to open models—for custom fine-tuning, low latency, or local deployment. Brad captures why we’re excited to partner with builders around the world to shape this new open model.
@romainhuet
-
OpenAI Partnership: Open Models for Custom Fine-tuning and Deployment
By
–
-
OpenAI to Release First Open Language Model Since GPT-2
By
–
You asked, we listened! We’ll release our first open language model since GPT‑2 in the coming months. Thanks for the great responses to @sama
’s poll! The results were close, so we’ll host events to gather feedback. We’d love your input to shape the model: https://
openai.com/open-model-fee
dback/
… -
Investigation into preset prompts tools and console errors
By
–
Thank you James! Do you have any specific tools turned on in your preset prompts, or any preset IDs / console errors to investigate? We’re looking into it—sorry for the trouble!
-
Semantic VAD improves speech-to-speech by detecting natural pauses
By
–
Building smooth speech-to-speech experiences can be tricky, especially handling natural pauses. Our new Semantic VAD in the Realtime API knows if someone’s still thinking or actually done talking, helping you avoid those awkward interruptions!
-
OpenAI.fm Text-to-Speech Interactive Demo Launched
By
–
http://
OpenAI.fm just dropped! We built an interactive text-to-speech demo for today’s livestream—and now it’s yours to play with. Type a script, pick your voice and vibe, hear it live—then grab a code snippet and start building! -
OpenAI announces exciting voice updates for developers
By
–
Another day, another surprise!
— Romain Huet (@romainhuet) 20 mars 2025
Tune in to @OpenAI at 10am PT—we’ve got some exciting updates for developers building with voice. 🎛️👀pic.twitter.com/1iUTNqAKESAnother day, another surprise! Tune in to @OpenAI at 10am PT—we’ve got some exciting updates for developers building with voice.
-
o1-pro API Now Available for Complex Agentic Tasks
By
–
By popular demand, o1-pro just landed in the API! Built for your toughest, most agentic problems—it uses extra compute to think harder and deliver consistently better answers.https://t.co/o0WAZImCd8pic.twitter.com/YnkM7uVwG0
— Romain Huet (@romainhuet) 19 mars 2025By popular demand, o1-pro just landed in the API! Built for your toughest, most agentic problems—it uses extra compute to think harder and deliver consistently better answers. https://
platform.openai.com/docs/models/o1
-pro
… -
PDF extraction with text and page images in context
By
–
It brings the extracted text and an image of each page of the PDF in the context. More details here:
-
PDF Content Extraction with Text and Images for Model Context
By
–
Yes! The extracted text and an image of each page are passed in the context to help the model understand PDF content.
-
OpenAI Responses API now supports PDFs for document processing
By
–
The @OpenAI Responses API now supports PDFs! Just pass a PDF, and it’ll expose its content directly to the model. Perfect for apps that extract info or summarize docs. Great working with @jaredpalmer and @lgrammel on bringing new features like this and web search to the AI SDK!