frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Show HN: Aphorio – 3,250,000 WebSocket sends in 0.002 seconds (Ryzen 9 9900X)

https://github.com/prettydiff/aphorio
1•austin-cheney•14s ago•0 comments

Math identity that applied at model-load time, makes MLA decode at 71,000 tok/s

https://efficientagent.substack.com/p/the-mla-decode-speedup-hiding-in
1•bnayak25•1m ago•0 comments

Putting Claude's Watermarking to the Test

https://dreasays.substack.com/p/putting-claudes-watermarking-to-the
1•gojkoa•1m ago•0 comments

Archer OS: draft spec for AI agents operating apps under OS authority

https://github.com/coachpato/archer-os
2•Billie_Archer•4m ago•0 comments

Remind HN: Perseid meteor shower peaks tonight

1•cjbarber•4m ago•0 comments

NIST RFI for modernizing the NVD in the wake of AI

https://www.federalregister.gov/documents/2026/08/12/2026-16371/request-for-information-rfi-on-mo...
2•MattSayar•5m ago•0 comments

Show HN: /show-me: agent skill for compact visual representations

https://www.humanlayer.com/blog/show-me-skill
1•dhorthy•5m ago•0 comments

Local LLM Hardware Calc

https://dubir.net/tools/local-llm-hardware-calculator/
1•delneg•5m ago•0 comments

QuestDB 10.0

https://questdb.com/blog/questdb-10-release/
1•tosh•6m ago•0 comments

Water Groups Push Washington for Cyber Rules After Hacking Spree

https://www.wsj.com/pro/cybersecurity/water-groups-push-washington-for-cyber-rules-after-hacking-...
1•toomuchtodo•7m ago•1 comments

SpaceXAI: Grok 4.6

https://openrouter.ai/x-ai/grok-4.6
2•theanonymousone•8m ago•0 comments

Samsung with Claude: "Design and verification shortened to 2 days from a month"

https://biz.chosun.com/it-science/ict/2026/08/12/XIEQWWZCDRFH7BJV5Z3DOY2RLQ/
1•TMWNN•10m ago•1 comments

Introducing Grok 4.6

https://cursor.com/blog/grok-4-6
3•stevefan1999•11m ago•0 comments

Hacker News TUI

https://github.com/alin9661/hnx
2•aaronlin098•11m ago•1 comments

Run OpenClaw with Muse Glimmer Locally on Mac

https://holaclaw.ai/docs/tutorials/use-muse-glimmer-with-openclaw
1•angelmm•11m ago•0 comments

How do payments work in the US, China, and Brazil?

https://andrewacomb.substack.com/p/a-tale-of-three-refunds
1•acombandrew•13m ago•0 comments

Apple Caps Bug Bounty Submissions After AI Surge

https://www.pcmag.com/news/apple-limits-bug-bounty-submissions-after-a-barrage-of-ai-entries
3•mempko•13m ago•0 comments

Researchers have successfully used AI to create brand new viruses

https://www.npr.org/2026/08/11/nx-s1-5927074/researchers-have-successfully-used-ai-to-create-bran...
2•Brajeshwar•14m ago•0 comments

Bounded Chaos

https://coldtake.dev/blog/bounded-chaos
1•AlarQ•15m ago•0 comments

Most of your tech debt is free

https://piechowski.io/post/most-of-your-tech-debt-is-free/
1•speckx•16m ago•0 comments

Show HN: GitHub Actions self-hosted runners on Modal Sandboxes

https://github.com/modal-projects/runner-modal
1•botirk•17m ago•1 comments

Show HN: Toposonico, a Music Map and Recommender

https://toposonico.com/#lon=9.9318&lat=-4.6522&z=14.00&entity=track&rowid=8714
1•deppep•21m ago•1 comments

Igalia and Bocoup: Cooperatives Unite

https://www.igalia.com/2026/JoiningForces.html
2•bkardell•22m ago•0 comments

Show HN: Learn to become a Rave Developer for the airline industry

https://olect.github.io/become-rave-developer/
1•olect•22m ago•1 comments

The security program that only works when everyone is at their desk

https://andreafortuna.org/2026/08/12/summer-cybersecurity-stress-test-nis2-dora/
1•speckx•24m ago•0 comments

Show HN: Seisin – a desktop app that turns your job search into analytics

https://getseisin.com
2•LudbaSH•24m ago•0 comments

Sharding a 70B model across 39 Intel laptops

https://github.com/labscommunity/cascadia
3•tatef•25m ago•1 comments

Apache Fory JSON: Fastest Java JSON Serialization, 10x Faster Than Jackson/Gson

https://fory.apache.org/blog/fory_json_fastest_java_json_framework/
1•chaokunyang•26m ago•1 comments

Grok 4.6

https://twitter.com/SpaceXAI/status/2087562800982077492
3•qingcharles•26m ago•1 comments

New Microsoft Defender 'ShieldBreak' zero-day grants SYSTEM privileges

https://www.bleepingcomputer.com/news/security/new-microsoft-defender-shieldbreak-zero-day-grants...
2•Brajeshwar•27m ago•1 comments
Open in hackernews

Qwen/Qwen3.8-2.4T-A95B

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B
81•Philpax•1h ago

Comments

volf_•35m ago
A ~5TB model.
PunchTornado•31m ago
The card looks almost too good to be true
guardiangod•30m ago
https://unsloth.ai/docs/models/qwen3.8

The 1bit quant model is at an astonishing 397GB with 95B active per MOE. This literally puts Opus 4.5 performance level into a machine a normal person could buy, and still gets usable tokens/second.

The full lossless model BF16 is clocking at 4.9TB. The model card claims the model to be between Opus 4.8 and Fable 5. Again that's astonishing as getting a machine with 7TB RAM (with context + KV cache) is still within the realm of medium size companies.

Bad things: The open source version has its vision capability removed, and the context capped at 250k . I expect someone to bolt a Kimi 2.6 vision tower to it to restore the vision capability (at less performance of course). For context, I played around with extending the context to 600k for Qwen 3.5 397b, and the context remained stable up to around 480k. It'd be interesting to see if the same can be done to Q3.8 .

Also no out of the box DSpark/DFlash support. MTP is present so we should at least get some boost in TP speed.

ilc•23m ago
To compare a 1 bit quant to the full fat model is misleading.

Honestly this model people at home can tinker with, if you have a big enough Mac. Maybe 4 Strix Halo/DGX Spark, and then at 1 bit quant? Nah.

Use the right sized model, for your hardware. You'll get better results.

guardiangod•21m ago
Extremely large 1 bit models are usually within 50-60% of KV divergence to lossless models. In this case I think the comparison to Opus 4.5 is a fair assessment.

Extremely large models don't suffer as much from quantization due to its weight topology also contains encoded information, so the loss of info from any one weight is somewhat mitigated.

ilc•14m ago
Any one weight, but all of them. And also crushing the architecture itself?

I wouldn't pick up 400gb of hardware to run in that mode. I might try it for fun, but even then you are looking at handling a 95GB active parameter set.

This is NOT a model for most home labs. I'm sure some can and will use it. But most, should steer clear.

auspiv•19m ago
Opus 4.5 level of performance is also accessible with deepseek-v4-flash-0731 (0731 being the july 31 update) which is much, much, much smaller. 2x RTX pro 6000 blackwell can run it. 4x can run it very comfortably
Philpax•15m ago
What do you need the extra 2 for? Tensor parallelism?
simonw•25m ago
Is this the largest ever open weight model release by parameter count? I think it is.
NitpickLawyer•23m ago
No, Kimi k3 is 2.8T params. This is 2.4T params but ~5TB weights because it was released in bf16 and ~2.5TB for the fp8 version. Kimi k3 launched with QAT 4bit, so ~1.5TB weights.
NitpickLawyer•25m ago
Supposedly this is a Kimi k3 rival. Bit of a chonker, especially since they only released bf16 and fp8. So at launch this will be harder to serve than k3. No QAT on q4 means that someone with deep pockets (nvda?) will have to quant it, with plenty of calibration data. Should bring it ~1.3TB, so around k3 size.

License pretty similar to k3 with some caveats. Free to use for internal or <50M$ revenue / year. Limitations above that threshold for serving the model or services targeting coding / productivity agents.

Benchmarks are looking good, trading blows w/ opus4.8 and sol, generally 10-20p under fable. But that's neither here nor there w/ qwen, their benchmark to real world usage correlation has been iffy in the past.

The local model 3.8-27B announced for Friday, same time so ~48 hours from now. That'll be a bit more exciting for a lot more people, since 3.6 was quite good for local inference, and their 3.7-max -> 3.8-max shows a lot of improvement.

MrDrMcCoy•10m ago
Llama.cpp can quantize without special training, but I'm not sure if any special model architecture support is needed to read it in the first place. If it can be converted to gguf at all and you know what tensors to target, it can get the full ternary bonsai treatment today.
binary132•4m ago
QAT is an optimizing quantization algorithm, not naive quant.
esafak•10m ago
They've got the price and performance right. I would like to see Chinese models improve their reasoning efficiency in the next generation.
l72•20m ago
> In particular, Qwen3.8-Max is the official version based on Qwen3.8-2.4T-A95B with more features, such as vision input & non-thinking support, 1M context length by default, official built-in tools, etc.

That is unfortunate, that the open weight model doesn't have vision support or the 1M context length...

wren6991•9m ago
People have had surprising success adding vision to open-weight LLMs that ship without it, like DSV4 Flash [1] or GLM-5.2 [2]. Given this model is already vision-trained I expect that approach will work well here.

[1] https://old.reddit.com/r/LocalLLaMA/comments/1vl6ior/i_gave_...

[2] https://huggingface.co/baseten/GLM-5.2-Vision-NVFP4

theanonymousone•10m ago
Do we know if AA and DeepSWE benchmarks are on bf16 or fp8 quantisations?
dhx•2m ago
Also of interest: DeepSeek V4-Pro-0813 (1.6T-A49B) benchmark scores have apparently just been announced on the DeepSeek WeChat channel and they're sitting about Fable 5 level.[1]

[1] https://www.reddit.com/r/LocalLLaMA/comments/1vmi0fg/deepsee...

arjie•11m ago
Longer context and more cache. The problem is that native format with DSpark enabled you have very little room on the VRAM.
guardiangod•13m ago
I am running DS v4 flash 0731 lossless at 80t/s right now. It really is not at Opus 4.5 level (for my workload). I would say it's around 3.7 Sonnet, which is still pretty good, but other models such as GLM 5.2 are still leaps better. Of course I run DSv4 flash over GLM 5.2 for a few very good reasons, but intelligence is not 1 of them.
MrDrMcCoy•6m ago
Despite fitting into VRAM, I can't get DSV4 to run at usable speeds on my AMD hardware. The upcoming qwen3.8 27b greatly excites me, and I hope it can outperform Stepfun 3.7 Flash, which is the best thing I can run today.