Hacker News Reader

_ache_on OpenDLSS: A Vulkan Reimplementation of Nvidia's DLSS 5 Neural Rendering Network

From the LICENSE

So... Either alooshdenny stole the commits, or it's an alias for maan.

Yes, that was my point. Good but too niche.

You point is valid for textual LLM, with large CoT, not Omni which will respond quick with a voice. In this case, token price is a good enough proxy.

What matter the most and isn't told by token price is the latency. You expect a voice LLM to respond very quick. If it takes 5s to response to a simple "Hello, what the weather today?", them not much people will use it.

_ache_on Qwen 3.8 Omni Flash

Very capable yes but very slow. 27B is relatively easy to run, but the 125b one need around 128Gb of RAM (DDR4 isn't enough, you need DDR5 to be quick enough, that's $3000 alone, you also need a graphic card). DDR4 is caped @20tps.

So, the cost of a setup to run Qwen-Flash-Next at +40tks is around $3000. Too much for most people.

With only a RTX 4090, you will reach 30tps (with DDR5...), not +40tks, and it's about the limit to be usable. Oh ! I forget Apple device too, it's a good option to run this model I guess, but still slow.

Yet, as you said, it's still a wip implementation, it may improve soon (MTP support is about to be merged in llama.cpp soon).

_ache_on Shapelearn Qwen 3.8 27B (13.1 GB VRAM)

Ahahah, thank you.

Yes, I'm not a native English speaker, and I'm definitely not an AI. ;)

I can't say the same thing because I just don't notice typos (mine or others).j I just assume my English is bad.

Isn't the definition of "being human" is "not to be perfect"? In French, it kinda is. We say "He/she humain after all" to mean that someone made a mistake.

_ache_on Shapelearn Qwen 3.8 27B (13.1 GB VRAM)

Ahahah :'D

Yes... I'm not english native. And definitively not an AI.

_ache_on Qwen 3.8 Omni Flash

I know, but it's a good enough proxy.

_ache_on Shapelearn Qwen 3.8 27B (13.1 GB VRAM)

From my own test. It's not faster than the unsloth model.

Disclarer: I'm unsing Vulkan on an AMD GC.

_ache_on Qwen 3.8 Omni Flash

If the performances are comparable, and there is no evidence it's not.

in/out ($) Gemini : 1.5 / 9.0 | Qwen 3.8: 0.15 / 0.47

That is a massive cost reduction.

Refs: https://www.alibabacloud.com/help/en/model-studio/model-pric... https://runware.ai/gemini-omni

_ache_on Qwen 3.8 Omni Flash

I don't think Qwen3.8-Omni-X will ever be released.

The last one was: Qwen3-Omni-30B-A3B https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct

And maybe Qwen4 won't be released, they only release Qwen3.8 27B (and a mostly unusable 125B). There are definitively slowing down open weight release.

_ache_on Mistral X Mozilla: Private, Multilingual AI Browsing

The strategies of Google and Apple, regarding how to provide a LLM, seam to disagree with you. Gemini run on a potato and Apple is local first.

So, you may actually have very good performance with local model. Just not yet on *every* device. So the Mozilla strategy here feel very reasonable. A Cloud provider specialised in local models, to be able to switch once local models will be quick enough on most devices.

_ache_on AI models don't kill people – people kill people

Ok, so OpenAI is going to compensate RubyGem for the cyberattack on its servers?

_ache_on GPT-6 Astra

https://ache.one/gpt6_now_down.png

Big claims, expensive and not release to the public yet.

_ache_on OpenAI begins rolling out GPT-6 Astra

It's up then down again. https://openai.com/index/gpt-6-astra/

What a bunch of amateurs. Here is it anyway :

https://ache.one/gpt6_now_down.png

The claims: https://share-md.com/view?id=870ba228-a25c-4169-bbc9-12d7f25...

And some others like this bugged Karts Game:

https://tidal-rush-paradise-gp.skirano.chatgpt.site/

This impressive spaceship construction game:

https://voidexplorer-shipyard.openai.chatgpt.site/?fleetSeed...

And a lot of graphs, some without even Astra on it. Oh and the logo is a Galaxy.

_ache_on I trained a small transformer in 1.5hrs and it beats many LLMs

In computer science, that is technically a language. A formal language if you want to look it up on Wikipedia.

_ache_on Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)

I will rephrase it. Will it run on any consumer hardware?

_ache_on Claude Code is going reduce limits by 25% from September 14

Reduce limits or usage? Twitter is blocked.

Can you do more or less?

_ache_on Show HN: See fiber breaks linked to a map

Is it linked to the last month COLT problem?

_ache_on Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)

I'm hearing Tencent, Zhipu and Baidu shaking from here. It's fair to assume BATX / 6 Tigers don't sleep very well either.

_ache_on Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)

I think a reasonable expectation of MAX requirement to claim "runable on consumer hardware" is to 32G VRAM and 128GB RAM and it run at +10tps.

_ache_on Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)

What will be the requirement, like 128G of RAM and 12G of VRAM ?

_ache_on Anthropic Risk August 2026 [pdf]

They need to train a new model every month to keep at the top of most benchmarks. They don't own any DC, the price is insane. Most of people are aiming at smaller models because Claude one's are too expansive. Evolution of intelligence of bigger models start to stagnate, smaller models are catching up.

27B local model just dropped, it's 6/8-month old SOTA. General ROI of AI investment is expected on a baseline of >10y.

_ache_on Qwen 3.8 27B

From your benchmark, Qwen3.8 is nearer than Opus 4.8 than Qwen3.6. 0.1pp but still.

Also, a lot of people don't really care about german language capacity, maybe people programming in DDP idk.

PS: You benchmark seems saturated. Most values sit @>75% in a benchmark generally indicate that it's no longer as useful as a <70% one. I mean, Qwen3.8 is 77.5% and Fable5 80%, the poll of values is from 65% to 90%.

_ache_on Anthropic Risk August 2026 [pdf]

I actually expect them to explain me how they will manage to not go bankrupt soon.

_ache_on Anthropic Risk August 2026 [pdf]

It's crazy how Anthropic talks so much about their "AGI risk" and not enough about the risk of bankruptcy.

_ache_on France's tax authority had data stolen on 680k taxpayers

It is already. You can buy it online. There is not a lot of places where hackers sell that kind of stuffs.

1k lines are already shared, seems legit. Most of them is just <30k€ people. Only a handful of millionaires (8 >10M if I remember correctly), 0 billionaires. 2/3 people, 1/3 pro, there is no information about tax of professionals.

Seems to be an API endpoint about people who ask questions to the DGFiP.

_ache_on GLM-5.3: Frontier coding with emergent cyber capabilities

No yet finished! Still waiting for tonight Qwen3.8-27B and the unsloth Q5_K_M/S quantification.

Hopping for an AgentWorld variant from Qwen but I guess, I have too high expectations.

_ache_on AI At Home Part 1: A Box Of Scraps

What can you do with "only" 64G of VRAM that a 32G can't? Also, the R9700 are so loud!

_ache_on Docker Sandboxes – Disposable, isolated sandboxes for AI agents

I planed to do exactly this, with podman instead of docker, volume support.

Like:

$ podman run -it --rm -v .:/workspace local-dev-ia /usr/bin/oc

Configured with a .env file. Hope to do it hopefully before the end of the week.

_ache_on Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

It is interesting but it does look like a careful distillation of (Spark and) biggers open-weight models.

The progress compared to Qwen3.6 27B is good, not that impressive, it's a 4 months old model. (kuto to them to compare to 27B dense and not 35B MoE, it's more fair to do so). It is very probable that Qwen3.8 27B will crush Glimmer-30B on most benchmarks.