pheggs

in my experience there's a huge difference in token usage when creating new projects versus working on existing ones

I know this is not entirely what this article is about, but I love software rendering for desktop apps. It seems to me that the performance impact with proper paint invalidation is negligible, but doesn't suffer from the same driver issues as hardware accelerated rendering seems to have. Bonus that it doesnt use any GPU resources in the age of AI

It could be.

should you not be worried more about sending data to providers of your own country? I think they are equally bad personally, but if you think one of them is acceptable, why do you prefer it to be your own government that has more ties to your life?

I have a different experience and find them just as good as the latest anthropic and openai models. But then, we probably can just have opinions on this as benchmarks are probably used for marketing to a large degree. I find it hard to find truly independent benchmarks that don't have any ties to the cash flow of openai and anthropic, can not be trained up on etc.

I find it curious that almost every AI post has way more comments than other posts, and I sometimes wonder how big of a part agents play here. I can't seem to distinguish whether it's just real interest in the subject or a big campaign trying to influence people's opinion

I dont buy that, mainly because in blind tests where people dont know which model was used and choose a result, this does not seem to hold true

I find it difficult to compare LLMs in general. It will do better in some cases, worse in others. As far as I see it, most of it is marketing. You can try it out extremely cheaply though through their API and OpenCode. I personally would recommend it.

I would not call them idiots personally, they are just facing an extraordinary difficult situation with their open weight competition that's almost impossible to fix. They will have to aim for the high-end B2B market eventually, and attempt to lock out their competition through regulation or other means

I was wrong, I thought they would only go for B2B already. But instead they will remove it from the subscription and price it at 50$/mtok. Compared to GLM-5.2 for 4.40$ that's quite a gap

Probably not. But I also cant imagine the hardware not improving. Also there is surely a sweet spot regarding price and quality. I have strong doubts that much larger models will give much better results, which is not like disk space where every additional bit makes it one bit better

thats true. also, I watched the glm prices and it didnt take long before the prices dropped even lower for some providers. its like another layer of competition between hosters

I am quite certain that it is delayed on purpose to maximize the gains, but at some point some company will see the huge demand for local ai and will want to eat the cake (given that it is feasible)