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

DeepSeek-V4-Flash Update

api-docs.deepseek.com

Post reply on HN
Print view View original

351–360 of 362 posts

Re: DeepSeek-V4-Flash Update

#351

Post by dotancohen »

addaon wrote: post #335
dotancohen wrote: post #321

Earlier quoted context omitted.

The only movie I've ever enjoyed after reading the book was Jurassic Park. I've never seen the Dune movies or most other book-based movies, but honestly The Martian and Project Hail Mary were tempting. I'll probably surrender to one of those two someday.

The most recent Dune movies pull off the incredible feat of being a better telling of the tale than the book, but only if you already know the tale from the books. They’re not beginner-friendly, but they’re extremely respectful to the book reader, while still make conscious (and in my opinion, positive) changes to the story.

Thank you. Maybe I'll read Dune again (or listen to an audiobook) then watch.

That sounds exactly like the type of experience I'd enjoy.

Re: DeepSeek-V4-Flash Update

#352

Post by ozgrakkurt »

siva7 wrote: post #232
dannyw wrote: post #91

Deepseek and moonshot are the only two providers I consent to training for.

CCP will be happy! Go on and share all your data with them..

CCP isn’t causing global economic crisis by bombing and invading foreign countries.

There is no telling how much worse US will get as they start to lose more and more internationally and grow the wealth inequality locally.

It is easy to be liberalist and cool when winning but it doesn’t mean there isn’t a racist, uneducated and violent majority

Re: DeepSeek-V4-Flash Update

#353

Post by MrDrMcCoy »

teravor wrote: post #340
Demiurge wrote: post #338

Earlier quoted context omitted.

Why would you do this?

a specialist model sufficiently post-trained can outperform a frontier model while being dirt cheap. what we did was distill GLM 5.2 into a 27B model on SQL and then post-train it with RL afterward. the result outperformed even Fable on that one task. the distillation step is just good sense in this workflow, to bootstrap a smaller model to the utmost you can before actually doing RL.

Do you have any reference guides on how to perform this distillation? I have a use case for this and hardware to run it on, but have never seen how it's done.

Re: DeepSeek-V4-Flash Update

#354

Post by anigbrowl »

hgoel wrote: post #350
anigbrowl wrote: post #291

Earlier quoted context omitted.

So go with some other provider who hosts the models like OpenRouter if are worried about this.

Some employers have a policy prohibiting any use of Chinese models, not even allowed to run them locally on air-gapped machines.

That's just paranoid, might as well argue that the models will sap an impurify all of our precious bodily fluids.

Re: DeepSeek-V4-Flash Update

#355

Post by econ »

johntarter wrote: post #306
flexagoon wrote: post #85

Earlier quoted context omitted.

Weird, I'm using 5.6 Sol through Chinese resellers and it reverse engineers stuff just fine

That's hilarious that Anthropic and OpenAI can't even secure their products from Chinese resellers, how are they supposed to secure their models from being distilled?

Same way you prevent them from crawling your public website?

Re: DeepSeek-V4-Flash Update

#356

Post by teravor »

MrDrMcCoy wrote: post #353
teravor wrote: post #340

Earlier quoted context omitted.

a specialist model sufficiently post-trained can outperform a frontier model while being dirt cheap. what we did was distill GLM 5.2 into a 27B model on SQL and then post-train it with RL afterward. the result outperformed even Fable on that one task. the distillation step is just good sense in this workflow, to bootstrap a smaller model to the utmost you can before actually doing RL.

Do you have any reference guides on how to perform this distillation? I have a use case for this and hardware to run it on, but have never seen how it's done.

nothing publicly available that I'm aware of.

Re: DeepSeek-V4-Flash Update

#357

Post by hgoel »

anigbrowl wrote: post #354
hgoel wrote: post #350

Earlier quoted context omitted.

Some employers have a policy prohibiting any use of Chinese models, not even allowed to run them locally on air-gapped machines.

That's just paranoid, might as well argue that the models will sap an impurify all of our precious bodily fluids.

I agree it's absurd. I was just commenting on the bit about organizations not allowing it.

Re: DeepSeek-V4-Flash Update

#358

Post by quicekuru »

f311a wrote: post #38

I've been driving flash model for 90% of my tasks. It's better than pro (for unknown reasons), very cheap and fast. I try to keep changes under 1000 lines and drive architectural decisions myself, barely notice any difference compared to frontier models. The rest 10% is to spot bugs, security problems and to investigate better architecture, which flash can also do pretty well, I just cross check it. Faster iterations…

Same experience here. Plugged it into my harness instead of a 10x more expensive model and it... just worked, sometimes even better than the expensive model in a direct comparison. Starting to think there is more value in speed than "intelligence" at this point.

Re: DeepSeek-V4-Flash Update

#359

Post by riskd »

1234letshaveatw wrote: post #214
riskd wrote: post #204

Earlier quoted context omitted.

Decades of xenophobic propaganda

Wow! Amazing observation! Wait, do you think there could have been any pro Chinese propaganda? Perhaps even some taking place at the present time? I think it could be possible?

You’re clearly too far gone to even fathom the thought; so you grasp at these straws for hope that the China you grew up “learning” about is the same in reality.

Re: DeepSeek-V4-Flash Update

#360

Post by Tepix »

zozbot234 wrote: post #239
Tepix wrote: post #217

Earlier quoted context omitted.

DS4Flash has 284B weights. 64GB? No go.

The native weights are 4-bit for the sparse experts, and they quantize to ~80GB with limited degradation in real-world performance. That's a viable target for 64GB with SSD streaming, though it will be slower than keeping the whole thing in RAM (especially on a M2-class machine with its slower storage).

I consider less than 4bit to be too poor quality usually. Got any benchmarks to measure quality?
Post reply on HN