Wanna see ppl fine-tuning it
DeepSeek V4 – almost on the frontier
131–140 of 420 posts
Re: DeepSeek V4 – almost on the frontier
#132Deepseek v4 Pro feels like Claude Opus 4.6 in it's personality but here's what I did find out about costs: I did cut loose Deepseek v4 on a decent sized Typescript codebase and asked it to only focus on a single endpoint and go in depth on it layer by layer (API, DTOs, service, database models) and form a complete picture of types involved and introduced and ensure no adhoc types are being introduced. It developed a…
I find a lot of the inefficiency also comes from the model just randomly poking around and grepping all the time which is the fault of the harness. I ended up building a Prolog based MCP where I use tree-sitter to parse the code into a graph, and then the model can just ask questions like 'what are all the functions connected to this function'. So, in case you're trying to focus on what a particular endpoint is doing, you can trivially and predictably trace the whole subgraphs of calls.
Re: DeepSeek V4 – almost on the frontier
#133The biggest differentiator for me: DeepSeek just does what I ask. I've tried using both GPT and Claude for reverse engineering recently, both refused. I even got a warning on my OpenAI account.
> I even got a warning on my OpenAI account.
This is kind of terrifying to me, regularly. No real manner of recourse to normal people without a following, potential exclusion from real fundamental tooling. Imagine OpenAI goes on to buy 20 companies and now you cant use Figma, Next, whatever just because you once tripped some very foggy line somehow. Not just OpenAI but the entire ecosystem is so... hard to read.
I was asking Gemini about a quote from catch 22 and it kept dying mid stream saying it cant talk about it, god knows why, it had no violent or sexual content -- though that is in the book. I could imagine it dinging my whole workspace account just because ... shrug?...
I know ideally the future is local, but I don't know how real that is for most people at least in the next few years with practical costs and power usage except I guess through a M* processor if you're in that ecosystem.
Re: DeepSeek V4 – almost on the frontier
#134Earlier quoted context omitted.
You definitely have a bone to pick. Chinese researchers usually have given the world the most cheap and consistent high quality research around LLMs. They don't pretend, they do the work and release the goodies. Mostly so cheap, every one in the world has a chance to use close to frontier models. Why would you respond with "Anger"? You let us know what your real complaint is about and let's not feign indignation at o…
You're making completely unfounded assumptions about me. I use Chinese models myself.
I made no such claims. Maybe you have something to share about why we need to have a negative view of free and open models based on publicly available frontier research.
Re: DeepSeek V4 – almost on the frontier
#135Jensen has a point. I believe these were trained and run on Huawei chips. The Nvidia embargo may backfire on American leadership as necessity gives way to invention.
Isn't it widely speculated that these are distilled from current frontier models? Distillation is far less compute intensive than primary training. That said, if distillation produces something almost as good for a fraction of the cost, Jensen's point may stand.
Re: DeepSeek V4 – almost on the frontier
#136Earlier quoted context omitted.
This is very false DS4 is super cheap. I would advise to begin by reading their release paper. https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main... They introduce very novel methods to improve long context efficiency and attention. HCA & mCH. It requires only 27% of flops for inference and 10% for KV cache than v3.2. This makes it super efficient. Think of this. For flops, we can now serve more than 3x the…
> I'm running this locally. Impressive! What is your setup? Are you running the full DeepSeek V4 Pro, or V4 Flash?
I'm running flash. You can run it under 128gb, so a $3000 strix halo would do. My rig tho is 8 Nvidia gpus and spilling over to system ram.
Re: DeepSeek V4 – almost on the frontier
#137Jensen has a point. I believe these were trained and run on Huawei chips. The Nvidia embargo may backfire on American leadership as necessity gives way to invention.
Isn't it widely speculated that these are distilled from current frontier models? Distillation is far less compute intensive than primary training. That said, if distillation produces something almost as good for a fraction of the cost, Jensen's point may stand.
You can't really distill a model without access to the internal weights. You could train on chat logs, but that's absolutely not the same thing, it doesn't even come close to comprehensively "extracting" the model's capabilities. And everyone does that in the industry anyway ever since ChatGPT was first released, some versions of Opus even claimed to be DeepSeek if you prompted them in Chinese.
Re: DeepSeek V4 – almost on the frontier
#138The biggest differentiator for me: DeepSeek just does what I ask. I've tried using both GPT and Claude for reverse engineering recently, both refused. I even got a warning on my OpenAI account.
We have an enterprise cursor account so I can try all the mainstream models. Using composer 2 on our own code which I obviously have the source code for I couldn't get it to turn on a debug flag to bypass license checks while I was troubleshooting something. Infuriating. It was like that old Patrick from SpongeBob meme.
I don't understand why we would turn the models into law enforcement officers. Things that are illegal are still illegal and we have professionals to deal with crimes. I don't need Google to be the arbiter of truth and justice. It's already bad enough trying to get accountability from law enforcement and they work for us.
Re: DeepSeek V4 – almost on the frontier
#139Deepseek v4 Pro feels like Claude Opus 4.6 in it's personality but here's what I did find out about costs: I did cut loose Deepseek v4 on a decent sized Typescript codebase and asked it to only focus on a single endpoint and go in depth on it layer by layer (API, DTOs, service, database models) and form a complete picture of types involved and introduced and ensure no adhoc types are being introduced. It developed a…
I've been having the same experience. Tasks like "go through this entire module and pedantically make it match my preferred styleguide exactly" were not worth a couple dollars with frontier models. It's nice to be able to put deepseek flash on stupid, unnecessary or highly speculative tasks without thinking about the cost.
Re: DeepSeek V4 – almost on the frontier
#140The biggest differentiator for me: DeepSeek just does what I ask. I've tried using both GPT and Claude for reverse engineering recently, both refused. I even got a warning on my OpenAI account.
We have an enterprise cursor account so I can try all the mainstream models. Using composer 2 on our own code which I obviously have the source code for I couldn't get it to turn on a debug flag to bypass license checks while I was troubleshooting something. Infuriating. It was like that old Patrick from SpongeBob meme. I don't understand why we would turn the models into law enforcement officers. Things that are ill…
They're probably worried about liability. Let's say that Oracle finds out you reverse engineered their DB using Gemini. You can be sure they will sue Google. Not just for providing the tools, but you could make the argument that it's actually Gemini doing the reverse engineering, and on Google's hardware no less.