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

DeepSeek V4 Flash 0731

arcprize.org

Post reply on HN
Print view View original

251–260 of 481 posts

Re: DeepSeek V4 Flash 0731

#251

Post by tsunamifury »

odo1242 wrote: post #236
anal_reactor wrote: post #222

Earlier quoted context omitted.

It's been true for almost every business. "Cheap and good enough" usually trumps "excellent but expensive". Ikea, McDonald's, Ryanair, AliExpress, Aldi - these brands prove that catering to poor people is more profitable than catering to rich people simply because there are so many poor people that their collective spending power outweights the one of rich people.

Well, not universally. It’s a tradeoff. If what you said was universally true Apple wouldn’t exist; Spirit Airlines wouldn’t be bankrupt, etc.

Apple sells to half the American population. And by definition many of them are poor.

Spirit was broken by oil prices which everyone pays the same for. (There is no cheaper jet fuel alternative).

Not a good comparison to the point of wrong conclusions.

Re: DeepSeek V4 Flash 0731

#252

Post by djc404 »

tarruda wrote: post #243

One of the best things about this version is that it is trained in the codex harness. It feels just as good as OpenAI models in using codex tools, but extremely cheap and with 1M context

Do you have any sense how using it with codex compares to OpenCode?

It’s always a bit tricky picking the right harness (when you have options). Sometimes the differences are subtle but meaningful. But who has the time to run everything twice and compare all the time!

Re: DeepSeek V4 Flash 0731

#253

Post by cyanydeez »

nomel wrote: post #198
cyanydeez wrote: post #128

Earlier quoted context omitted.

If I had the GPU size, hook it up to llama.cpp and setup the --reasoning-budget and reasoning-message; Most of that additional reasoning is a lot of garbage and you can redirect it to useful output. That's how I handle the Qwen27B and 35B

> Most of that additional reasoning is a lot of garbage and you can redirect it to useful output. What do you mean by "redirect it to useful output"? Could you give an example? This sounds interesting.

It's specific to the harness. Using dynamic context pruning, the budget cuts it off after a select amount of tokens and the budget message tells the model to use subgents to finish whatever it's thinking about

Re: DeepSeek V4 Flash 0731

#256

Post by stockworks »

horsh1 wrote: post #228
abixb wrote: post #162

Earlier quoted context omitted.

If what you're saying is true and accurate, then US-based AI labs are in big trouble. The only saving grace might be some sort of a 'national security' proclamation banning the use of state-of-the-art Chinese (and non-US) models across US federal and state governments and large enterprises (especially ones with federal government contracts), but even still, US AI labs will probably lose out massively on international…

If programming in the US to become unconditionally 10x more expensive, then the exodus from the US is about to begin.

I couldn't agree more, and think of all the wasted inference across accounts overpaying for their subscriptions.. Need a secondary marketplace for this stuff.

Re: DeepSeek V4 Flash 0731

#257

Post by segfault99 »

kennywinker wrote: post #28
minimaxir wrote: post #16

Earlier quoted context omitted.

DeepSeek V4 Flash 0731 is an open-weights model which means price is determined by competition/invisible hand of the marketplace: https://openrouter.ai/deepseek/deepseek-v4-flash-0731 With the exception of cache costs, all providers have similar input/output costs.

Not counting the cost of making the model, which is subsidized by… someone? The chinese gov i think?

DS comes out (one of, or) the most successful quant fund in China.

They don't strictly need any kind of subsidies.

FWIW they have a funding round planned (kerfuffle about leaks from CEO presentation few weeks back) -- presumably because infrastructure needs have ballooned.

Naturally there will be some PRC government interest in one of their flagship AI companies. From what is visible seems to be more along the lines of ensuring that DS gets its fair share of resources -- e.g. Xi Jinping meeting founder and positive comments about success of DS means that (hypothetically) Alibaba can't screw DS too much on infra charges to kill off a 'competitor'. Also would imagine that DS's top guys have been clearly identified and will have been 'discouraged' from going to work for one of the SV polycules. But even here as much carrot as stick -- none of the DS top guys will ever need to work again except for love of the job.

Re: DeepSeek V4 Flash 0731

#258

Post by nomel »

cyanydeez wrote: post #253
nomel wrote: post #198

Earlier quoted context omitted.

> Most of that additional reasoning is a lot of garbage and you can redirect it to useful output. What do you mean by "redirect it to useful output"? Could you give an example? This sounds interesting.

It's specific to the harness. Using dynamic context pruning, the budget cuts it off after a select amount of tokens and the budget message tells the model to use subgents to finish whatever it's thinking about

Nice. Does it use a summarization, or a hard cutoff?

Re: DeepSeek V4 Flash 0731

#259

Post by stiltzkin »

_s_a_m_ wrote: post #93
LaurensBER wrote: post #9

I've been using it extensively since the release and the best summary I can give is that it's good enough to use it for (almost) everything and cheap enough that the cost are irrelevant. I'm running it in Oh My Pi with a second instance running as "advisor" and even with 5-6 active sessions (effectively 12 streams) I'm struggling to spend more than 5 bucks per day. OpenCode Go even has double limits temporarily so fo…

These posts have to be Chinese bots, these models are all trash. Used it via OpenCode for an hour, cost me one hour of my life. It is for anything complete trash.

[dead]

Re: DeepSeek V4 Flash 0731

#260

Post by jboss10 »

LaurensBER wrote: post #220
paxys wrote: post #118

Earlier quoted context omitted.

How is $5/day irrelevant? In the $150/mo range you can get effectively unlimited usage of GPT 5.6 Sol (Pro plan). Why use a much weaker model for the same price?

5 USD is at the "raw" API price. OpenCode currently offers 60 USD API credits at 10 USD per month (OpenCode Go) and have even doubled it temporarily as a promotion. Effectively you can get Deepseek for 1/12th the already ridiculous cheap API price.

From here, it looks like opencode is hemorrhaging money. I've got a Opencode Zen free account, and I've been using deepseek-v4-flash-free on Pi for a bit, and I haven't hit a limit yet. Sometimes my request fails, but retrys work. I know this is a very cheap model, but it's being given out for free. I assume they might be training on outputs?
Post reply on HN