DeepSeek v4.1 Flash • Hacker News Board

Re: DeepSeek v4.1 Flash

#531

Post by stymaar » Thu, Sep 10, 2026, 9:15 PM UTC

ungovernableCat wrote: post #199

rao-v wrote: post #20

As I also said on Twitter - it really amazes me how fearless Deepseek are. Every single model release is packed with new and crazy clever ideas and somehow, they always commit to training them at near frontier scale. I know everybody wants the tell all story of the clever ideas that were developed over the last ~3 years at Anthropic and OpenAI, but what I really want to thumb through is DeepSeek's notebook of "brilli…

Its CEO allegedly holds a 84% stake and he's the same guy who founded the hedge fund that funds it. Deep pockets + simple control = perfect culture to just hire talent and let them go wild without worrying about financial viability, as long as the king CEO is fine with it that is

Unless that CEO is Mark Zuckerberg, I guess.

Re: DeepSeek v4.1 Flash

#532

Post by pampas » Thu, Sep 10, 2026, 9:31 PM UTC

gandreani wrote: post #397

pampas wrote: post #161

I've run some evals on my puzzle game https://redactle.net/llm-leaderboard Deepseek v4.1 flash is able to solve it some of the time. I've found it burns through more reasoning tokens than any other model. Google models like Gemini 3.8 Flash are still dominating and is able to one-shot most evals while being the cheapest. I'm curious what other unique evals people are running.

It's so bizarre having a low score be GOOD. It's like reverse intuition. Shouldn't it be called `score error` or something along those lines?

Great point. I've changed the naming.

Re: DeepSeek v4.1 Flash

#533

Post by ValentineC » Thu, Sep 10, 2026, 9:34 PM UTC

mentalgear wrote: post #105

https://xcancel.com/deepseek_ai/status/2097930608790167907 Should be the link ( now that it works again! :) )

I wish AI companies wouldn't post their primary announcements on fElon-enshittified Twitter.

Use Bluesky or, I don't know, have a news site. They could vibecode one in minutes.

Re: DeepSeek v4.1 Flash

#534

Post by kkukshtel » Thu, Sep 10, 2026, 9:35 PM UTC

ungovernableCat wrote: post #199

rao-v wrote: post #20

As I also said on Twitter - it really amazes me how fearless Deepseek are. Every single model release is packed with new and crazy clever ideas and somehow, they always commit to training them at near frontier scale. I know everybody wants the tell all story of the clever ideas that were developed over the last ~3 years at Anthropic and OpenAI, but what I really want to thumb through is DeepSeek's notebook of "brilli…

Its CEO allegedly holds a 84% stake and he's the same guy who founded the hedge fund that funds it. Deep pockets + simple control = perfect culture to just hire talent and let them go wild without worrying about financial viability, as long as the king CEO is fine with it that is

This is also what is compelling about Midjourney imo.

Re: DeepSeek v4.1 Flash

#535

Post by nottorp » Thu, Sep 10, 2026, 9:45 PM UTC

muddi900 wrote: post #449

nottorp wrote: post #206

Earlier quoted context omitted.

Anthropic and OpenAI will threaten you every 3-6 weeks. It's their marketing strategy. It's too bad because the tools can actually be useful. If you consider them tools.

A stick is the most basic of tools. A stick is also the most basic of weapons.

Heard of the boy who cried wolf?

They have so many dangerous breakthroughs per year that by the time they actually have a breakthrough no one's going to even read the press release...