Artificial Intelligence thread

[iewgnem]

How do they make money at the end of the day? Since being so cheap(almost free) gotta incur some costs no? Else all Chinese models will be as cheap as deepseek.

Deepseek has always been very focused on efficency over capability, they invented a lot of optimization techniques and Huawei's inference chip were designed with their input, plus theres a good chnce they're using research from their back pocket they havent published yet.

And lets not forget they're a hedge fund, so they can probably make good money timing the AI market lol

 

[siegecrossbow]

Deepseek has always been very focused on efficency over capability, they invented a lot of optimization techniques and Huawei's inference chip were designed with their input, plus theres a good chnce they're using research from their back pocket they havent published yet.

And lets not forget they're a hedge fund, so they can probably make good money timing the AI market lol

The only kind of shorting I support — greedy fucks think they can send stocks to the moon by gutting white collar workforce and sending everyone to trade school.

 

[Michael90]

Deepseek has always been very focused on efficency over capability, they invented a lot of optimization techniques and Huawei's inference chip were designed with their input, plus theres a good chnce they're using research from their back pocket they havent published yet.

And lets not forget they're a hedge fund, so they can probably make good money timing the AI market lol

Is that not kind of illegal in China ? These types of short sellers often disrupt markets like in the west.
anyway, im just asking if their model is sustainable in the long run(say a few years from now) if they can keep it up or will they change down the road once they have accumulated enough market share compared to their more expensive Chinese rivals ?

 

[tphuang]

GLM works great on OpenCode. Zcode is great too but I prefer CLI, if Zai come out with a CLI I might switch.

I wouldnt touch Claude Code with a 10ft pole and that was true even when I was using Claude

yeah, everyone can find their own combination on whatever it works. I'm glad the Chinese govt is finally taking this Claude code threat seriously.

Thats what I mean by people dont actually use it, the company pay bulk price for everyone and people who use less average out ones who use more, which means the system breaks down if everyone actually start using it at high rates, which is probably what MS, Uber and Coinbase found out.

I'm sitting on around 1B GLM5.2 tokens on my dashboard for last 30 days and I have my team sharing my $100/mo Kimi plan, on GLM I only came close to 5h limit once and I dont even think about rate on Kimi. So Im sitting at a team of 4 people doing enterprise dev without even checking tokens on $250/mo.

So as I said, the only reason anyone would use Claude is if they dont know the bill

Ps. These are only for AI at around Opus level that I need for coding, for general purpose, task automation, doc summery, etc, Deepseek is for all practical purposes free, I still havent worked through my initial $50

which Zai plan are you using? I'm basically using the cheapest GLM and Kimi plan. For a one person operation, it's enough. did hit my one week limit on Zai once, so I'm now saving the easy work for Kimi. But some of the easiest task should really be done with a normal editor. The thing that annoys me the most about ZCode is that it doesn't allow me to edit files.

People literally can't use it fast enough because we are all producing way more code than we can review and approve without blindly trusting Claude to not take down our systems, which would cost the company 1000x more than what we get from shipping code faster. Not like it's bad at reviewing, mind you, we just can't afford to make mistakes on production. The limiting factor is human attention, not AI budgets.

Just big corporate things.

yeah, using Claude really just enabled a lot of check-ins without actual productive gains.

IMO the one I am waiting for is Deepseek V4 release and if it's GLM 5.2-tier. If they manage to do it, and keep the costs reasonably low along with their cache pricing, it's all over.

Imagine having semi-frontier level coding intelligence for 0.2 cents or whatever, per million cached input tokens lol

so, it's going to be 3 months since V4 preview came out. if they cannot get to more or less 5.2 tier on this release, that's a very big disappointment. Because GLM will supposedly come out with 5.5 next month. And based on what we saw with 5.1 to 5.2, this 5.5 should be quite the jump in performance.

 

[tphuang]

https://twitter.com/i/web/status/2075276608693321841

Ant Group Ling Team with LingBot-Video model. This is quite cool. Although after Seedance 2.0, it's hard to get the same excitement over video models.