Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Qwen3.8-Max: A New Bar for Coding and Cowork

qwen.ai

Post reply on HN
Print view View original

611–620 of 652 posts

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#611

Post by jpgvm »

niea_11 wrote: post #499
jpgvm wrote: post #375

Earlier quoted context omitted.

They can be deterministic. We did this at Groq, if you sent a request with exactly the same input token, seed and temperature value you would get precisely the same result every time. This is harder to do on other architectures that themselves aren't fully deterministic though.

Do you mean that you get the exact same result (as word for word) because from Groq's documentation, it says that determinism is best effort, and also it's not guaranteed across model versions : https://console.groq.com/docs/prompting?_sm_vck=tMW2S7k17245... My understanding is that it's best effort because even if you control the seed/temperature, the model version and hardware specs you still have floating point ma…

> harder to do on other architectures that themselves aren't fully deterministic though

The TLDR here is that Groq LPUs are completely deterministic hardware, including for floating point.

Disclaimer: Formerly of Groq, now of NVIDIA, still working on LPUs.

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#612

Post by seanmcdirmid »

alightsoul wrote: post #527
seanmcdirmid wrote: post #312

Earlier quoted context omitted.

Water. The west is really dry. We could throw in data lines as well but it’s really bottlenecked by water. Actually the only place in China with water in abundance is southeast China. Also they really don’t have the chips to supply those data centers, and the government isn’t so crazy to go that aggressive on AI, they have lots of other priorities as well.

Chinese data centers like to use air cooled chillers for cooling, not evaporative coolers. It's simple because power for the chillers is cheaper than water for the evaporative coolers. The opposite is true in the US

[flagged]

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#613

Post by dgellow »

AbsurdCensor wrote: post #603
dgellow wrote: post #572

Earlier quoted context omitted.

And it would be in seconds if models weren’t that large and slow-ish to download! LM studio is such a noob friendly experience, pretty neat first experience!

At least for the most part, if you are downloading from huggingface, you should be able to saturate your connection. I know I usually can pretty easily even with a 5gig connection at home.

Hmm, let’s not talk about my German poor internet connection please :)

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#614

Post by seanmcdirmid »

HarHarVeryFunny wrote: post #488
seanmcdirmid wrote: post #312

Earlier quoted context omitted.

Water. The west is really dry. We could throw in data lines as well but it’s really bottlenecked by water. Actually the only place in China with water in abundance is southeast China. Also they really don’t have the chips to supply those data centers, and the government isn’t so crazy to go that aggressive on AI, they have lots of other priorities as well.

One of China's big sources of power is hydroelectric such as the Three Gorges Dam, so if you want electricity and water in one place, then there it is.

Try doing evaporative cooling in a swamp and tell me how that works.

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#615

Post by ttflee »

toshinoriyagi wrote: post #8

They've also announced Qwen3.8-27B being released open-weight next week. Qwen3.6-27B is widely regarded as one of the best local models, especially since nothing else comes close to it, that isn't benchmaxxed, without being significantly larger. If 3.8 truly improves upon it that would be awesome.

It is really a difficult choice in front of me: DeepSeek v4 flash 0731@q2 vs Qwen 3.{6,8} 27B@fp8 on a 96GB VRAM server.

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#616

Post by senordevnyc »

anon373839 wrote: post #590
senordevnyc wrote: post #482

Earlier quoted context omitted.

For example, swindling their business partners (such as Figma). Filing an unnecessary trademark lawsuit against a customer whose brand predates Anthropic. Completely unrelated to AI safety. they literally sell the removal of guardrails in exchange for minimum spend commitments It seems blindingly obvious to me that you'd want to let trusted organizations use the models defensively, which means without guardrails, and…

1. This is not “unrelated to AI safety”. The things they seek in the name of AI safety always seem to be aligned with the path that will eliminate competition and give them maximum economic power. Thus, to disentangle what is sincere and what is ulterior, we have to look at the leadership’s character holistically. A ruthlessly unethical and cutthroat corporation that is also seeking unprecedented power doesn’t deserv…

You're just talking in circles. So their trademark lawsuit is about AI safety because of the leadership's holistic character?

Again, you just have an axe to grind, and you've picked a poor area in which to grind it.

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#617

Post by anon373839 »

senordevnyc wrote: post #616
anon373839 wrote: post #590

Earlier quoted context omitted.

1. This is not “unrelated to AI safety”. The things they seek in the name of AI safety always seem to be aligned with the path that will eliminate competition and give them maximum economic power. Thus, to disentangle what is sincere and what is ulterior, we have to look at the leadership’s character holistically. A ruthlessly unethical and cutthroat corporation that is also seeking unprecedented power doesn’t deserv…

You're just talking in circles. So their trademark lawsuit is about AI safety because of the leadership's holistic character? Again, you just have an axe to grind, and you've picked a poor area in which to grind it.

Let me try putting this in simpler terms:

When someone says they are doing something to help you, but it really helps them, you have to choose if you trust what they say. One way to know if you can trust them is to watch how they behave.

(By the way, I have been very transparent on HN that I have an axe to grind, literally referring to it as “an axe to grind”. My axe is that I think the frontier labs are a symptom of a cancer on our society, and I personally wish to see them and their tactics fail.)

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#618

Post by HarHarVeryFunny »

seanmcdirmid wrote: post #614
HarHarVeryFunny wrote: post #488

Earlier quoted context omitted.

One of China's big sources of power is hydroelectric such as the Three Gorges Dam, so if you want electricity and water in one place, then there it is.

Try doing evaporative cooling in a swamp and tell me how that works.

So then don't do that - the US trend is towards closed-loop cooling.

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#619

Post by sourav276 »

trollbridge wrote: post #149
nozzlegear wrote: post #27

Earlier quoted context omitted.

Qwen3.6-35B is my daily driver for AI, and what convinced me to cancel my Claude subscription back in April. The Qwen3.6 line is easily the best local model I've tried, and I've tried a lot. I've got it diligently grinding away on my laptop right now, reviewing and fixing some bugs in my F# code.

Qwen-3.6-35B-A3B was our "gateway drug" into switching our organisation to agent/harness-first coding. Particularly, I had one team member who was extremely sceptical of AIs/LLMs/harnesses and refused to use them. One day he said "Well, I have an RTX 5090 doing nothing... should I try to get something up on it?" and a few minutes later he had 3.6-35B loaded up, running OpenCode. It continues to be a workhorse to this…

[flagged]

Re: Qwen3.8-Max: A New Bar for Coding and Cowork

#620

Post by ahartmetz »

jiggawatts wrote: post #578
ahartmetz wrote: post #266

Earlier quoted context omitted.

It took me about three hours total to set up a local model. I already have a GPU and I have fiber for the download. llama.cpp is not difficult to compile and has many backends. It can run parts of the model on different backends, like in the common case that the GPU doesn't have enough VRAM for everything. There are many step-by-step guides available.

Three hours is a lot longer than one minute.

No shit, but the huge ordeal you described is an exaggeration.
Post reply on HN