It's fascinating how well o1 generalizes across different subjects.

Test time compute drastically increases compute demand. Even beyond scaling, which is already massive.

scaling works

Btw this is how the intelligence explosion begins

This is a pretty big deal. We're rapidly moving towards a world where you can decide how many GPU-hours you want to dedicate to a problem. More GPU-hours = more thinking = better solution.

gg.

“this isn’t a one-off improvement – it’s a new scaling paradigm and we’re just getting started.” September 12, 2024 — the day the growth curve turned skyward to the Singularity and the intelligence explosion truly began.

Awesome work.

Some reflection on what today's reasoning launch really means: New Paradigm I really hope people understand that this is a new paradigm: don't expect the same pace, schedule, or dynamics of pre-training era. I believe the rate of improvement on evals with our reasoning models

I just got access

Hope you guys have strapped your seatbelts.

rollout complete; live to 100% of chatgpt plus/team users now

Intelligence will be cheap & plentiful. Progress will be impactful & lasting. And the world will be joyful & prosperous. The AI Abundance Age has begun.

This release is much closer to the original GPT-3 release than the ChatGPT release. A new paradigm, which some will find incredibly valuable for things even we at OpenAI can’t predict. But it’s not a mass product that just works and unlocks new value for everyone effortlessly.

The reasoning core doesn’t need to be big. Most parameters are used for knowledge and memorization. Factor them out!

Every major lab has been saying openly for the last six months that we are not at the top of the S-curve, and that they can see nothing but air above us. As it turns out they were telling the truth. The question now is, how much further can we go?

I just checked the front pages of some mainstream news websites and there’s no mention of what happened today. I’m not shocked but this really is so bizarre. The vast majority of the world’s population is completely in the dark

remember anthropic's claim that 2025-2026 the gap is going to be too large for any competitor to catch up? that o1 rl data fly wheel is escape velocity

Remember this, from the May 28th announcement? I’m wondering whether OpenAI was referring to o1 or some other model

near Sam Altman o1-mini is a dinner roll on the table, while o1-preview is a bread loaf in the oven. the loaf smells good, but it’s still baking. if you’re hungry for math now, snack on the dinner roll.