Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

developers.openai.com

Post reply on HN
Print view View original

211–220 of 352 posts

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#211

Post by hadlock »

colingauvin wrote: post #55
eigenspace wrote: post #9

The fact that AI models can be so easily distilled and replicated is such a stroke of luck. 10 or 15 years ago if one had asked me to envision a future where a private company invents artificial intelligence, I'd have thought for sure they'd have a massive moat, be very difficult to catch, and it would create an almost instant monopoly. Rather, it seems that selling intelligence might end up as a race to the bottom.…

Where is the actual evidence of distillation? I keep seeing this repeated ad nauseam but I must have somehow missed the evidence.

It turns out you can train a 1b model at almost 1000 tokens/s on a m5 max laptop. As a personal experiment, I've been asking Sol for synthetic training data and synthetic agentic training data (model distillation in it's purest form), plus modified opencode, codex transcripts etc for training data, and nobody's even paying me to do it. If I'm doing it has a hobby, you can bet industrial users are doing it.

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#212

Post by porridgeraisin »

voxic11 wrote: post #61
colingauvin wrote: post #55

Earlier quoted context omitted.

Where is the actual evidence of distillation? I keep seeing this repeated ad nauseam but I must have somehow missed the evidence.

Distillation a pretty well documented technique that actually pre-dates LLMs https://arxiv.org/pdf/1503.02531 Here is a project that guides you through it if you want to prove to yourself that it works https://github.com/arcee-ai/DistillKit

To be precise, the distillation mentioned in this paper is not the distillation used by other model companies. In the one mentioned in the paper, your teacher and student model typically have similar architectures - and you typically need access to the full logits. What happens here instead is motivated by the fact that these companies don't have access to the training data and compute that anthropic/openai have. The distillation they do basically amounts to using traces from ant/oai models trained on much more data with a lot more compute (in many cases including the hidden intermediate tokens! turns out there were many ways to coax it out) and then either directly training on it or using it in many ways in post training pipelines. It falls under imitation learning, IMO.

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#213

Post by Vetch »

kingstnap wrote: post #117
adastra22 wrote: post #95

Earlier quoted context omitted.

This viewpoint doesn't make any sense to me. The weights + inference code are the "source code" for AI. I literally don't know what else you are demanding for the "open source" label.

If you think of LLMs as programs. The weights and inference code are very much a binary. While the training code and data are the true source. Since if you want to robustly modify the LLM that's actually what you need. But since "compilation" (training) is extremely compute intensive this isn't something accessible to anyone without an entire datacenter. Anyway semantics aside having the binary is still infinitely be…

The weights + the architecture are already 100% of the code, the transformer is just a mathematical expression + helper programs whose sources are provided. The transformer itself is not even a stateful program, so a it is no more a binary than Piet or Tromp's BLC are. It's merely incomprehensible. Training isn't compilation either, since training a model is closer to program induction and the data are samples defining the solution space.

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#214

Post by adastra22 »

kingstnap wrote: post #117
adastra22 wrote: post #95

Earlier quoted context omitted.

This viewpoint doesn't make any sense to me. The weights + inference code are the "source code" for AI. I literally don't know what else you are demanding for the "open source" label.

If you think of LLMs as programs. The weights and inference code are very much a binary. While the training code and data are the true source. Since if you want to robustly modify the LLM that's actually what you need. But since "compilation" (training) is extremely compute intensive this isn't something accessible to anyone without an entire datacenter. Anyway semantics aside having the binary is still infinitely be…

That’s not how it works though. Two training runs on the same data don’t produce the same weights. And if you want to modify the AI, you do so by fine tuning the weights not rerunning training. In every respect that matters, the weights are both the binary and the source code together.

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#215

Post by ceejayoz »

rootusrootus wrote: post #87
ninjahawk1 wrote: post #25

Once they make a model better than Fable I’ll be switching to Codex. Their priorities in terms of consumers seem to be better. I do think Anthropic has some solid safety viewpoints, but I don’t necessarily think that either is entirely aligned yet with delivering exactly what humanity needs. Maybe the AI will help align the AI companies when it gets smart enough. That’s the real misalignment I’m concerned about.

It feels like 5.6-Sol is already fairly close to Fable, and in some ways exceeds it. Just the other day I had Fable draw up a solution for me, and then I fed it into 5.6-Sol and said how does this look ... it found an oversight and told me about it, and when I then fed that observation back into Claude it acknowledged the miss. I've noticed also that 5.6-Sol is more concise with output than Fable (and let's not talk…

> Just the other day I had Fable draw up a solution for me, and then I fed it into 5.6-Sol and said how does this look…

You should be doing this for every solution.

Even Fable reviewing itself will find issues, unproven assertions, etc. Same for Codex models. A review loop is critical.

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#216

Post by endymi0n »

colingauvin wrote: post #55
eigenspace wrote: post #9

The fact that AI models can be so easily distilled and replicated is such a stroke of luck. 10 or 15 years ago if one had asked me to envision a future where a private company invents artificial intelligence, I'd have thought for sure they'd have a massive moat, be very difficult to catch, and it would create an almost instant monopoly. Rather, it seems that selling intelligence might end up as a race to the bottom.…

Where is the actual evidence of distillation? I keep seeing this repeated ad nauseam but I must have somehow missed the evidence.

Been using a lot of Kimi K3 lately and the answers have been… „load-bearing“ to the point of hilariousness. It‘s obvious from where they distilled, even if sceptics rightly point out it can‘t have been the only source of their secret sauce, as it‘s been better than the current Opus 4.x at the time of release.

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#217

Post by Razengan »

eigenspace wrote: post #9

The fact that AI models can be so easily distilled and replicated is such a stroke of luck. 10 or 15 years ago if one had asked me to envision a future where a private company invents artificial intelligence, I'd have thought for sure they'd have a massive moat, be very difficult to catch, and it would create an almost instant monopoly. Rather, it seems that selling intelligence might end up as a race to the bottom.…

Was it not obvious that the value and advantage was going to be in AI-adjacent services?

The quality of the harness UX, and random fun crap like Sora, it's a shame that OpenAI killed that so soon, and also Group Chats in ChatGPT.. they risk running a Googlelike reputation at this rate

Maybe ultimately whomever can be the "Apple of AI" will win

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#218

Post by calumcl »

wahnfrieden wrote: post #201
calumcl wrote: post #147

Earlier quoted context omitted.

Subscribers were already getting subsidised compute and value compared to the 20-200$ fee, peak cakeism to want more considering the alternative would likely be consumption based pricing for individuals so you "benefit" from OpenAI giving up some of their markup (and heavy users end up SOL).

Looks like you’ve confused clarification for begging

There was context for neither in your original post and I personally saw a lot of similar takes like the above online - apologies for the misunderstanding. I personally thought it was pretty clear considering the parent link is a link to the API pricing and their public communication on their twitter/blogs were "Subscription usage remains unchanged".

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#219

Post by esafak »

Culonavirus wrote: post #145
rsyring wrote: post #108

Earlier quoted context omitted.

It's common for different models to find holes in another's work. There are various good reasons for that. FWIW, we use ChatGPT for our primary model and use Claude to do the reviews. This works better than ChatGPT doing it's own review even with a clean session/context.

Same here. Grok Build 4.6 for me, given how cheap Grok is and how Sol is supposed to be "the" SOTA, it finds a surprising amount of bugs. Most of which Sol agrees with needs to be fixed or improved. I've done this tens of times between these two models and it works great in my experience. Sol initial back and forth with me. Commit. Let Grok review. Sol fix. Only then do I start reading the code.

I suspect it would work with the models swapped too, or even with one model and a blank context for the second run.

Re: OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)

#220

Post by ipsod »

rsyring wrote: post #108
rootusrootus wrote: post #87

Earlier quoted context omitted.

It feels like 5.6-Sol is already fairly close to Fable, and in some ways exceeds it. Just the other day I had Fable draw up a solution for me, and then I fed it into 5.6-Sol and said how does this look ... it found an oversight and told me about it, and when I then fed that observation back into Claude it acknowledged the miss. I've noticed also that 5.6-Sol is more concise with output than Fable (and let's not talk…

It's common for different models to find holes in another's work. There are various good reasons for that. FWIW, we use ChatGPT for our primary model and use Claude to do the reviews. This works better than ChatGPT doing it's own review even with a clean session/context.

It's beyond common for a model to find holes in its own work, as well. I have an iterative review as the part of all agentic work, and it always finds something to fix, and will sometimes spend hours fixing its own work.
Post reply on HN