undefined

points

[-]

1M cached tokens on deepseek is $0.006, the big labs can't sell anywhere close to this, they have funders expecting returns and huge overhead.

btw I've had a ton of fun with the new deepseek today, I was waiting for my OpenAI 5h limit reset and decided to give it some problems for fun, got pretty great results. Tried some harder problems and still got great results. I don't expect it to be Sol class or anything but I really didn't expect it to be anywhere near this good so we'll see where it ends up. And it's really fun throwing crazy amount of tokens at the wall for ~free instead of watching the subscription limits tick closer while your agents churn away.

by eru8 hours ago|

parent|

[-]

> 1M cached tokens on deepseek is $0.006, the big labs can't sell anywhere close to this, they have funders expecting returns and huge overhead.

Losing most of your customers tends to sharpen the mind a bit. They could eg stop pushing out the absolute frontier for a while and focus on making what they have run cheaper. Or they go and do more lobbying against China. Or a million other little things that take more than 30 seconds to come up with when writing a HN comment, but less than a week for someone who's smart and paid to do this for a living.

by eru8 hours ago|

parent|

prev|

[-]

I did a lot of that kind of work with the older version of DeepSeek before they upped the prices.

For example, it was quite good to get a decent Sashiko review. Sashiko is a Linux kernel review agent with interchangeable LLM driver. It's very good, but it eats tokens like crazy.

by user4392813 hours ago|

parent|

prev|

[-]

That's 1/3 of GPT 5.6 Luna. It seems rather close to me.

But great that we have a new leader in performance/price in that segment.

by cmrdporcupine13 hours ago|

parent|

prev|

[-]

Yep. 4.1 Flash is good enough for most routine coding things, but it also makes up for a lot of weakness by being so fast (and cheap of course).

I'm willing to tolerate babysitting things a lot more if I know I'll get almost instant results.

by eru8 hours ago|

parent|

[-]

You could also have eg Sol do the babysitting.

by colingauvin15 hours ago|

prev|

[-]

OpenAI just paused new subscriptions to their $200 plan. They are in a rock and a hard place. Obviously the Astras and Fables of the world are exponentially more expensive, but for...less than exponential returns. The question is whether they can leverage the marginal advantage into something that justifies the diminishing returns before the bottom catches up to them.

On the one hand you, if you bought a lot of compute a couple years ago (perceived demand, perceived shortage) you are in a good spot temporarily. But the counter to that is that everyone else is becoming more compute efficient so maybe that advantage isn't what people thought it would be. I can almost, almost run DS4.1 Flash at home. 4 sparks can do it at 200+ tokens per second. I have two Sparks, so I am not in the club. Neither is your average laptop owner or gamer either. But your average HN software engineer can probably easily swing 2 sparks.

by eru8 hours ago|

parent|

[-]

> But the counter to that is that everyone else is becoming more compute efficient so maybe that advantage isn't what people thought it would be.

I'm not sure? If we have techniques to use the hardware even better, that will make the hardware even more valuable, won't it?

by colingauvin6 hours ago|

parent|

[-]

But they aren't really competitive for cost in that size class.

by user4392813 hours ago|

parent|

prev|

[-]

I could buy four of them for ~20k.

That's like four years of ChatGPT + Claude subscription.

Eight years if only ChatGPT, or sixteen years of the Pro 5x subscription.

by colingauvin9 hours ago|

parent|

[-]

It's not particularly good value if you are just comparing $ with no other context. Point is just that it's accessible and so now Anthropic and OpenAI need to make both a performance proposition and a value proposition.