segmondy

12 ms·

331.
by segmondy 7mo ago
folks have too much money than sense, gpt-oss-120b full quant runs on my quad 3090 at 100tk/sec and that's with llama.cpp, with vllm it will probably run at 150tk/sec and that's without batching.
332.
by segmondy 7mo ago
it's for fools. i bought 160gb of vram for $1000 last year. 96gb of p40 VRAM can be had for under $1000. And it will run gpt-oss-120b Q8 at probably 30tk/sec
333.
by segmondy 7mo ago
combination of both, they published papers so we can clearly see they are not just duplicating old methods but coming up with new optimizations. ... yet we can't rule out that they used Nvidia. I don't even see how the export r
334.
by segmondy 7mo ago
what chaff? Just search, find what you want and buy. It's like ebay.
335.
by segmondy 7mo ago
Well, it's not the same HP. If there was ever a case that Ship of Theseus is not the same it's with companies. It just takes but a few replacement to get an entirely different company, mostly same people, same name, same busines
336.
by segmondy 7mo ago
This is very false. My first system was a 3060 which you can buy new for about $300 or used for about $200. If you already have an existing system you can use it, else you can pick up a used PC for about $150. Entry is about $500.
337.
by segmondy 7mo ago
vibe one. ;-)
338.
by segmondy 7mo ago
Everyone outside the US doesn't deal with USD. Your comment is bonkers. Read up on purchasing power. All locations are not equal.
339.
by segmondy 7mo ago
open source alternative, at first it's going to suck. but over time it will win. imagine how miserable we would be if all we had was windows and osx. but we have linux. we are now at such crossroads were the choice is android an
340.
by segmondy 7mo ago
Opencode is not delusional, it would be delusional to think Anthropic won't after they have already threatened them.
341.
by segmondy 7mo ago
so they own the trademark to Clawd? Or is it suppose to be Claude?
342.
by segmondy 7mo ago
Can you show any service that is selling API because they have turned a $200/mo Claude code account into an API service? Give me a break.
343.
by segmondy 7mo ago
Anthropic is a shit company. I cancelled my subscription 2 years ago once they started calling for regulation. They might have gotten folks to side with them in the OpenAI debate, but they are just another shit company like OpenAI.
344.
by segmondy 7mo ago
I was wondering the same, but llama.cpp was written to offload to system ram. If this really works, then the advantage could be that one could run transformers / sglang, etc or other tools that don't offload to system ram. Howev
345.
by segmondy 7mo ago
They don't owe anyone an answer.
346.
by segmondy 7mo ago
Ha, keep putting your prompts and workflows into cloud models. They are not okay with being a platform, they intend to cannibalize all businesses. Quality doesn't always reign over control and trust. Your data and original ideas are
347.
by segmondy 7mo ago
... and then a raspberry pi today can outdo all of those computers.
348.
by segmondy 7mo ago
I use llama.vim - https://github.com/ggml-org/llama.vim I run my own models, you can have models for code suggestion/FIM in the middle and you can have another for chat mode. You can point it to your local model
349.
by segmondy 8mo ago
Which quant? I find folks running lower quants complaining, yet they should be running higher quant. Qwen3CoderNext is great, even at Q6. I mistakenly had it loaded for an agentic workflow and was surprised at how well it is.
350.
by segmondy 8mo ago
I see lots of people defending this. What if the owner doesn't post, but reads and uses DM? What if they post the delete their posts when it gets old? Like Michael Burry?
351.
by segmondy 8mo ago
nah, if AI dream comes to fruition, then we can rewrite all the code and system to be better. I mean, we are pretty wasteful from resource perspective. We would need 1/20th the CPUs and ram. Your intel 586 cpu will be as good as y
352.
by segmondy 8mo ago
well, not if this AI stuff pans out. AI will be reversing engineering and making things easy. ;-)
353.
by segmondy 8mo ago
I hope Larry screws up big time and blows up Oracle so it can go the way of Sun.
354.
by segmondy 8mo ago
AI probably flagged it or someone made a bogus request. I have seen this happen numerous times, it takes folks having to fight it to get their accounts back. If the person has no strength or interest in fighting it as in this case, then t
355.
by segmondy 8mo ago
You can replace Sonnet and Opus with local models, you just need to run the larger ones.
356.
by segmondy 8mo ago
It's truly an amazing model from the small models all the way to 397B. I wish they had released one as a FIM model.
357.
by segmondy 8mo ago
What's amusing is that developers who are the most technical competent instead of figuring out how to run these models locally are using cloud models and putting in their code and workflow in there, helping to train these models and dr
358.
by segmondy 8mo ago
You have no idea what you are talking about if you are asking the model what size it is or claiming that a model lies.
359.
by segmondy 8mo ago
I don't know that this is a first model release. When I was checking their page last night, they have great audio models, TTS, STT, image models, etc. I'm skeptical that folks do all of that on the first release. Possible but u
360.
by segmondy 8mo ago
Folks are now optimizing for promotion and recognition. So there are endless thirsty developers ready to tear down a stable software so they can be recognized as the one that rebuilt it and made it "better" either for promotion

More ›