Deepseek R1-0528 • Hacker News Board

Re: Deepseek R1-0528

#11

Post by hadlock » Wed, May 28, 2025, 6:37 PM UTC

transcriptase wrote: post #2

Out of sheer curiosity: What’s required for the average Joe to use this, even at a glacial pace, in terms of hardware? Or is it even possible without using smart person magic to append enchanted numbers and make it smaller for us masses?

As mentioned you can run this on a server board with 768+ gb memory in cpu mode. Average joe is going to be running quantized 30b (not 600b+) models on an $300/$400/$900 8/12/16gb GPU

Re: Deepseek R1-0528

#12

Post by hd4 » Wed, May 28, 2025, 6:41 PM UTC

willchen wrote: post #4

I love how Deepseek just casually drops new updates (that deliver big improvements) without fanfare.

On the day Nvidia report earnings too. Pretty sure it's just a coincidence, bro.

Re: Deepseek R1-0528

#13

Post by rahimnathwani » Wed, May 28, 2025, 6:44 PM UTC

hadlock wrote: post #11

transcriptase wrote: post #2

Out of sheer curiosity: What’s required for the average Joe to use this, even at a glacial pace, in terms of hardware? Or is it even possible without using smart person magic to append enchanted numbers and make it smaller for us masses?

As mentioned you can run this on a server board with 768+ gb memory in cpu mode. Average joe is going to be running quantized 30b (not 600b+) models on an $300/$400/$900 8/12/16gb GPU

I'm not sure that's enough RAM to run it at full precision (FP8).

This guy ran a 4-bit quantized version with 768GB RAM: https://news.ycombinator.com/item?id=42897205

Re: Deepseek R1-0528

#14

Post by latchkey » Wed, May 28, 2025, 6:52 PM UTC

hu3 wrote: post #9

transcriptase wrote: post #2

Out of sheer curiosity: What’s required for the average Joe to use this, even at a glacial pace, in terms of hardware? Or is it even possible without using smart person magic to append enchanted numbers and make it smaller for us masses?

It's probably going to be free at OpenRouter. There's already a 685B parameter DeepSeek V3 for free there. https://openrouter.ai/deepseek/deepseek-chat-v3-0324:free

It is free to use, but you're feeding OR data and someone is profiting off that.

Re: Deepseek R1-0528

#15

Post by SkyPuncher » Wed, May 28, 2025, 6:54 PM UTC

transcriptase wrote: post #2

Out of sheer curiosity: What’s required for the average Joe to use this, even at a glacial pace, in terms of hardware? Or is it even possible without using smart person magic to append enchanted numbers and make it smaller for us masses?

Practically, smaller, quantized versions of R1 can be run on a pretty typically Macbook Pro setup. Quantized versions are definitely less performant, but they will absolutely run.

Truthfully, it's just not worth it. You either run these things so slowly that you're wasting your time or you have to buy 4- or 5-figures of hardware that's going to sit, mostly unused.

Re: Deepseek R1-0528

#16

Post by doctoboggan » Wed, May 28, 2025, 7:01 PM UTC

willchen wrote: post #4

I love how Deepseek just casually drops new updates (that deliver big improvements) without fanfare.

Honest question, how do you know this is a big improvement? Are there any benchmarks anywhere?

Re: Deepseek R1-0528

#17

Post by ankit219 » Wed, May 28, 2025, 7:02 PM UTC

latchkey wrote: post #14

hu3 wrote: post #9

Earlier quoted context omitted.

It's probably going to be free at OpenRouter. There's already a 685B parameter DeepSeek V3 for free there. https://openrouter.ai/deepseek/deepseek-chat-v3-0324:free

It is free to use, but you're feeding OR data and someone is profiting off that.

Thats how a lot of application layer startups are going to make money. There is a bunch of high quality usage data. Either you monetize it yourself (cursor), get acquired (windsurf) or provide that data to others at a fee (lmsys, mercor). This is inevitable and a market for this is just going to increaase. If you want to prevent this as an org, there arent many ways out. Either use open source models you can deploy, or deal directly with model providers where you can sign specific contracts.

Re: Deepseek R1-0528

#18

Post by margorczynski » Wed, May 28, 2025, 7:03 PM UTC

hd4 wrote: post #12

willchen wrote: post #4

I love how Deepseek just casually drops new updates (that deliver big improvements) without fanfare.

On the day Nvidia report earnings too. Pretty sure it's just a coincidence, bro.

Yeah the timing seems strange. Considering how much money will move hands based on those results this might be some kind of play to manipulate the market at least a bit.

Re: Deepseek R1-0528

#19

Post by modeless » Wed, May 28, 2025, 7:04 PM UTC

willchen wrote: post #4

I love how Deepseek just casually drops new updates (that deliver big improvements) without fanfare.

I like it too, but some benchmark numbers would be nice at least.

Re: Deepseek R1-0528

#20

Post by danielhanchen » Wed, May 28, 2025, 7:05 PM UTC

transcriptase wrote: post #2

Out of sheer curiosity: What’s required for the average Joe to use this, even at a glacial pace, in terms of hardware? Or is it even possible without using smart person magic to append enchanted numbers and make it smaller for us masses?

We made DeepSeek R1 run on a local device via offloading and 1.58bit quantization :) https://unsloth.ai/blog/deepseekr1-dynamic

I'm working on the new one!