Hongyu Ren
@ren_hongyu
Followers
15K
Following
501
Statuses
200
Research Scientist @openai. CS PhD @stanford. Lead the development of o3-mini and o1-mini.
Stanford, CA
Joined April 2018
o3-mini is here! Together with @shengjia_zhao, @_kevinlu, @max_a_schwarzer, @ericmitchellai, @brian_zq, @sandersted and many others, we trained this efficient reasoning model, maximally compressing the intelligence from big brothers o1 / o3. The model is very good in hard math/coding/science questions with a fraction of cost and latency, defining new cost-efficient reasoning frontier. The model supports three different reasoning powers. Users can adjust the thinking time based on different use cases. The longer the model thinks, the better its capability is. With o3-mini-low, we drastically reduce latency compared to o1-mini, achieving GPT-4o level latency for response. You can apply early access to this model today to do safety-testing!
55
145
1K
Humanity's last exam gets doubled in two days from o3-mini to deep research, nice progress! 🚀🚀🚀
Very excited to finally share OpenAI's "deep research" model, which achieves twice the score of o3-mini on Humanity's Last Exam, and can even perform some tasks that would take PhD experts 10+ hours to do! A few thoughts on the implications: Deep research can be seen as a new interface for the internet, in addition to being an incredible agent. The current way that we humans use the internet is quite clunky--it takes time to search and click around, we're limited by time and attention, and we don't know the full range of websites out there. AI, on the other hand, never gets tired, can browse many websites at a time and has nearly unlimited world knowledge. With a deep research model that browses the web so well, we are entering a world where any information that would take a human hours to compile, can be synthesized by AI for you in a few minutes. We now have the luxury of: - Instead of combing through hundreds of reviews to decide on a hotel, restaurant, or product, AI will synthesize all the relevant info for us - When we travel, we'll get the best tips from all corners of the web, even ones from niche forums - For any new scientific idea, we can quickly get a list of all the relevant literature, related ideas, and people to connect with It's basically like a version of the internet personalized for what you want to know. This paradigm will be so powerful that in the future, navigating the internet manually via a browser will be "old-school", like performing arithmetic calculations by hand instead of using a calculator.
10
10
208
We released o3-mini today! Everyone can use it for free. It reasons hard, reasons fast, searches the web, and most importantly, knows research. Ask the model hard questions and brainstorm with it!
46
32
761
I will be on Reddit to answer your questions on o3-mini in 15min!
Reddit AMA with @sama, @markchen90, @kevinweil, @snsf, @michpokrass, and @ren_hongyu going live soon.
16
4
207
RT @sama: o3-mini is out! smart, fast model. available in ChatGPT and API. it can search the web, and it shows its thinking. available…
0
2K
0
RT @_aidan_clark_: o3-mini's intelligence x speed combo is incredible, idk what to say other than just try it and see for yourself. This t…
0
49
0
RT @SebastienBubeck: o3-mini is a remarkable model. Somehow it has *grokked arxiv* in a way that no other model on the planet has, turning…
0
116
0
I have been playing R1 for a while, it's very impressive in Chinese. Congrats to the @deepseek_ai team!
🚀 DeepSeek-R1 is here! ⚡ Performance on par with OpenAI-o1 📖 Fully open-source model & technical report 🏆 MIT licensed: Distill & commercialize freely! 🌐 Website & API are live now! Try DeepThink at today! 🐋 1/n
12
11
198
RT @sama: big news: the free tier of chatgpt is going to get o3-mini! (and the plus tier will get tons of o3-mini usage)
0
2K
0
RT @_aidan_clark_: o3-mini first try no edits, took 20 sec (told me how to convert to gif too.....) Get excited :)
0
89
0
RT @OpenAI: Announcing The Stargate Project The Stargate Project is a new company which intends to invest $500 billion over the next four…
0
12K
0
it's a good model. safe, smart and fast
thank you to the external safety researchers who tested o3-mini. we have now finalized a version and are beginning the release process; planning to ship in ~a couple of weeks. also, we heard the feedback: will launch api and chatgpt at the same time! (it's very good.)
23
9
482
Join the group to work with all of us to break the data wall, push the frontier on both intelligence and efficiency, and build the next generation of agent framework. It’s a lot of fun working on this research, and land it as a beneficial product for everyone.
if you’d enjoy pushing the frontier in AI, definitely apply for this job:
4
7
103
RT @mckbrando: to anyone wondering if the high ARC-AGI score is due to how we prompt the model: nah. I wrote down a prompt format that I th…
0
88
0
one of the best thread on o3 👇
o3 represents enormous progress in general-domain reasoning with RL — excited that we were able to announce some results today! Here’s a summary of what we shared about o3 in the livestream (1/n)
2
3
77
RT @__nmca__: o3 represents enormous progress in general-domain reasoning with RL — excited that we were able to announce some results toda…
0
383
0