frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Steam Machine launches today

https://store.steampowered.com/news/group/45479024/view/685257114654870245
1073•theschwa•6h ago•955 comments

Unsloth GLM-5.2 – How to Run Locally

https://unsloth.ai/docs/models/glm-5.2
93•TechTechTech•2h ago•31 comments

Optocam Zero: a Pi Zero based digital camera made using off the shelf components

https://github.com/dorukkumkumoglu/optocamzero
75•iamnothere•4h ago•20 comments

British Columbia, Time Zones, and Postgres

https://www.crunchydata.com/blog/british-columbia-and-time-zone-changes
91•sprawl_•4h ago•54 comments

My Mathematical Regression

https://blog.dahl.dev/posts/my-mathematical-regression/
173•aleda145•3d ago•59 comments

Japanese symbols that speak without words

https://arun.is/blog/japan-symbols/
78•msephton•4h ago•19 comments

Moebius: 0.2B image inpainting model with 10B-level performance

https://hustvl.github.io/Moebius/
214•DSemba•9h ago•60 comments

Canada plans 'nuclear renaissance' with up to 10 reactors built by 2040

https://www.cbc.ca/news/politics/federal-nuclear-strategy-9.7244509
216•geox•4h ago•102 comments

Show HN: Oak – Git alternative designed for agents

https://oak.space/oak/oak
129•zdgeier•8h ago•128 comments

Is it time for a new Embedded Linux build system?

https://yoebuild.org/blog/time-for-a-new-build-system/
16•cbrake•4d ago•6 comments

Canyon HUD helmet for road riding

https://media-centre.canyon.com/en-INT/266866-new-canyon-heads-up-display-helmet-could-be-a-safet...
48•zh3•2d ago•54 comments

Kyber (YC W23) Is Hiring a Head of Engineering

https://www.ycombinator.com/companies/kyber/jobs/FGmI8mx-head-of-engineering
1•asontha•2h ago

Flock-Powered Police Chiefs Stalking Women Shows Why Warrants Are Needed

https://ipvm.com/reports/police-chiefs-track
253•jhonovich•4h ago•88 comments

PivCo-Huffman "Merge" Operations

https://fgiesen.wordpress.com/2026/06/21/pivco-huffman-merge-operations/
26•luu•1d ago•0 comments

Nintendo Wii U games running from a 1980's Bernoulli disk [video]

https://www.youtube.com/watch?v=8GZDOpV2OXk
84•zdw•1d ago•35 comments

DisplayMate

https://www.displaymate.com/
85•skibz•7h ago•26 comments

Nearly half of LG smart TV apps contain residential proxy SDKs

https://spur.us/blog/smart-tv-apps-residential-proxy-sdks
169•microcode•2h ago•106 comments

Finding the Best Dog Treat with Statistics

https://www.wespiser.com/posts/2026-06-19-best-dog-treat.html
75•wespiser_2018•6h ago•19 comments

Linux and Secure Boot certificate expiration (2025)

https://lwn.net/Articles/1029767/
92•weaksauce•5h ago•48 comments

Cyberdecks, going analog, and convivial technology

https://blog.hydroponictrash.solar/cyberdecks-going-analog-and-convivial-technology/
3•akkartik•2d ago•0 comments

Pledging another $400k to the Zig software foundation

https://mitchellh.com/writing/zig-donation-2026
707•tosh•10h ago•236 comments

Die analysis of the 8087 math coprocessor's fast bit shifter (2020)

https://www.righto.com/2020/05/die-analysis-of-8087-math-coprocessors.html
77•Jimmc414•10h ago•16 comments

Job application asked for my SAT scores

https://mrmarket.lol/job-application-asked-for-my-sat-scores/
40•seltzerboys•2h ago•91 comments

Chevron signs 20-year power agreement with Microsoft for West Texas data center

https://www.chevron.com/newsroom/2026/q2/chevron-signs-20-year-power-agreement-with-microsoft-for...
111•cdrnsf•10h ago•103 comments

The text in Claude Code’s “Extended Thinking” output

https://patrickmccanna.net/the-text-in-claude-codes-extended-thinking-output-is-not-authentic/
263•0o_MrPatrick_o0•9h ago•184 comments

Deno Desktop

https://docs.deno.com/runtime/desktop/
1008•GeneralMaximus•18h ago•368 comments

Memory crisis is getting so bad that even retro RAM prices are going to the Moon

https://www.theregister.com/personal-tech/2026/06/22/the-memory-crisis-is-getting-so-bad-that-eve...
88•speckx•4h ago•25 comments

Help I accidentally a wigglegram

https://lmao.center/blog/wiggle-accidents/
471•gregsadetsky•2d ago•119 comments

Show HN: Got sick of ads, so I made my own logic puzzle site

https://puzzlelair.com/
123•HaxleRose•11h ago•94 comments

Blogger defeats photographer's copyright claim

https://blog.ericgoldman.org/archives/2026/06/blogger-defeats-photographers-copyright-claim-sokol...
87•speckx•6h ago•53 comments
Open in hackernews

Unsloth GLM-5.2 – How to Run Locally

https://unsloth.ai/docs/models/glm-5.2
92•TechTechTech•2h ago

Comments

xrd•1h ago
So close! My machine with 192GB RAM + RTX 3090 24GB can almost run this. It says it needs 24GB of VRAM and 256GB of RAM for MoE offloading.

https://unsloth.ai/docs/models/glm-5.2#usage-guide

In a prior thread, someone said it would take $500k in hardware:

https://news.ycombinator.com/item?id=48629970

mgambati•1h ago
With 2 wouldn’t have good results. Ideal range for coding is at least Q8.
kibibu•1h ago
According to this very article, 4-bit dynamic is essentially lossless
cheema33•1h ago
I have the RAM, but not the VRAM. What kind of speed/tps could you expect from a 3090 with 24GBs of RAM? I am somewhat tempted to pick a GPU with 24GBs of RAM.
elliotbnvl•36m ago
$500k is a vast overestimation. For massive concurrency at FP8 or even BF16 maybe.

NVFP4 at reasonable speeds (~120 tok/s) and concurrency is possible at a $80/90k figure with today's prices, maybe even less. That buys you 6 RTX 6000 PRO Blackwells, a decent CPU and motherboard, power supply. 576gb of VRAM.

You could do it for under $50k if you're OK with 40 tok/s decode, ~1200 tok/s prefill.

zuzululu•1h ago
wonder if AMD's new ai chip can run this with ease? I'm seriously consider buying it. GLM 5.2 is just shy of GPT 5.4 so I would welcome offloading any grunt work locally

I am very excited for local LLMs I think we may have GPT 5.5-xhigh level of performance for under 2000 EUR

This should put more pressure on the frontier models to avoid sitting on any fancy stuff and lower token prices as a whole.

Nothing beats a local LLM disconnected from the cloud.

Iolaum•1h ago
At full quantization GLM 5.2 may be close to GPT 5.4. But at Q2 or whatever one needs in order to run it on a pro-sumer device it will be worse.

Also I m not sure where you are getting the under 2k value. I bought a Framework desktop 128GB last year and my setup was around 2.7k. The same setup now sells for around 4.7k.

kccqzy•58m ago
The AMD 395 supports up to 128GB unified RAM. So still not enough even at 1-bit quant unfortunately.
benjiro29•53m ago
"GLM 5.2 is just shy of GPT 5.4"... If your running the full model. As in have 750 (FP8) to 1.5TB(FP16) of memory available.

Do not mix the benchmark results of GLM 5.2 FP16/FP8 with FP4 or FP2.

* FP4 will mean a accuracy loss of about 3%. Not noticeable but more chance for mistakes.

* FP2 ... what is what most people are able to run at home, for a "reasonable" price. Your looking at over 17% loss in accuracy.

At that point, your running at less then claude-sonnet-4.6, as the issues compound with accuracy losses. And reasonable priced is still in the ~ $5000 range (192GB + GPU 32GB active/kv cache system).

For that price your using a Codex / Claude Pro subscription for the next 4+ years with better models (by default), let alone with a FP2 GLM 5.2 version. And your looking at < 10 fps. A MacStudio with 512GB will net you 18 a 20fps+ with FP4, but ... i mean, those used to be $10.000.

Unfortunately the local hardware cost is a major issue for running large models like that.

pheggs•1h ago
I feel like the gap is closing to be able to run good enough models locally even for coding and I would assume it could make some companies a bit nervous. Am I wrong about that?
fny•1h ago
The RAM requirements are still pretty painful.
yieldcrv•49m ago
equilibrium in one or two more years on the consumer/prosumer side

think Apple M6 or M7 with a currently unforeseen denser memory style, 256gb RAM

a couple inference or cache improvements on the algorithmic side, using less ram for context windows and doubling token speed again

denser open source models, packing more experts for smaller active layers

it'll still be expensive but like $8,000 - $13,000 instead of $450,000 worth of B200s

stingraycharles•30m ago
Fairly certain that model sizes and computational requirements will grow as the price for LLM compute drops.
CamouflagedKiwi•1h ago
The hardware requirements to run this locally are still very high. Seems far enough off mainstream for those companies not to be too worried yet.
cogman10•59m ago
I don't think so. I could easily see a company deciding to host and run these models for their own development. If you have a dev team of about 10 people, a one time $50k investment in an LLM server has to be pretty tempting. Unlimited tokens, decent performance, upgrade options, and potential product integrations.

For companies wanting LLMs in their products in general, I have to think going the local llm route is even more tempting. Somewhat dumb models are more than good enough for a lot of the things people are integrating LLMs into their products.

andai•36m ago
How is this model half the size of DeepSeek V4 Pro? Is it because DeepSeek did more aggressive cost cutting on the attention mechanism?
skiing_crawling•20m ago
"it can fit" on 256GB of RAM, but it will be heavily quantized and still run very slowly. The headline number is not token generation, its prompt processing. So if you get 10 tok/s and an API gives you 20-30 tok/s, it doesn't seem that bad on its face, but a mac studio or any other machine that's not loading all of it into GPU will do PP 20-50X slower than a purely GPU based setup, which is what actually makes this unusable without $50k in GPUs.

On top of that, you will still be heavily quantized.

nullc•19m ago
Just running cpu only w/ Q6 on 9684X I get about 1tok/s ... also still get about 1tok/s/stream when running 16 in parallel.
Wowfunhappy•13m ago
> The full model requires 1.51TB of disk space

...a bit of an odd question: how well do LLMs losslessly compress, as in for cold storage?

I definitely don't have the hardware to run this model at any kind of reasonable speed (and I don't want to use a super aggressive quantization that would kill performance). Even so, I think it would be cool to retain an offline copy, in case... I don't really know, a solar flare destroys the internet some day, or maybe a zombie apocalypse. It would just be cool to have.

But 1.5 TB is a bit too much! If it could be compressed down into something semi kind of reasonable, that would be fun!

nh43215rgb•49m ago
Even with upcoming AI Max+ PRO 495 we are capped with 192GB, so no...
nl•27m ago
> I am very excited for local LLMs I think we may have GPT 5.5-xhigh level of performance for under 2000 EUR

We are maybe 10 years off that.

RAM prices are going to continue to increase for the next 2 years at least.

Even putting that aside it's currently around 40-70,000 EUR to run this with a FP8 quantization (which you need to get close to maximum performance).

To actually get GPT 5.5-xhigh performance in the real world you need more headroom to support things like subagents (which will fill up your KV cache).

I like local models but realism is important. The sweet spot for the next 3 years will continue to be ~35B MoE models. They might match GPT 5.5-xhigh for chat-style problems but not for coding.

hsuduebc2•20m ago
I wonder, if in the near future any acquisitions of some RAM producers with intent to just keep RAM prices up, will happen from the AI companies. It could seriously hurt their business, if companies will be able to host their AI in some time.
UncleOxidant•19m ago
Are you talking about Medusa Halo? It's going to support up to 256GB unified memory (up from 128GB for Strix Halo and 192GB for Gorgon Halo). That might just be barely enough to run a 2-bit quant GLM-5.2. It will expand memory bus to 384-bits, vs. 256-bits for Strix Halo which will help with bandwidth (projected to be around 500 GB/sec). But don't expect Madusa Halo-based machines to appear until sometime in 2028.

The other way this could go is that Z.ai could decide to release a smaller model targeted towards coding. They've done that before (GLM-4.7-Flash had 30B params). It would be great if they decided to release something in the 80B-100B param range. Something that size would easily run in a current Strix Halo system.

eventualcomp•35m ago
Where is $50k coming from again?
stingraycharles•32m ago
That’s less than the monthly salary of 10 software engineers, and assuming they pay API prices, probably earns itself back in about a year.

Having said that, I don’t think it’s all that tempting for companies at all, considering this whole market is developing rapidly and it’s nearly impossible to predict where we’ll be at in a year or two.

cogman10•21m ago
The hardware requirements aren't evolving and the local models have only been improving.

It's not like you'd lose capabilities, if anything this solution just gets better with time.

cogman10•27m ago
As in who pays for it or how did I arrive at that number?

For who pays for it, obviously the employer would.

For "how did I arrive at this number" Ballpark estimate from what I know about part cost. Most of that money will go towards AI cards about $5k for the mb, cpu, power supply, etc. $45k would be for as much ram and as big/expensive nVidia cards as you can get your hands on. The B300 has 288GB of VRAM in it. Probably what you'd be after.

twelvechairs•20m ago
Surely for most the desire is just an LLM provider that doesnt store or sell their queries (including by national actors). As long as that is allowed to happen surely its the answer for the vast majority.
UncleOxidant•28m ago
If we didn't have a RAM/GPU shortage right now they would be more nervous than they are. But as it is very few people are going to be able to afford a rig that can run this model effectively. That's probably not going to change for several more years yet. I think if the Z.ai folks decide to come out with a flash version of GLM-5.2 specialized for coding that came in about about 80B params, then the US frontier labs would probably be more worried. Overall, the Chinese AI companies have been showing the way to do the same amount with less (sometimes much less) and as that trend continues it's going to make the frontier labs worried - but even the Chinese AI companies are going to want to protect their moat by not releasing capable models that are significantly smaller than their current flagship models. AliBaba Qwen seems to be there now - it's gotten mighty quiet from them lately - their latest 395B model is just too large for most folks to run at home and they don't seem to be making any noises about releasing smaller ones this time around.