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

Qwen3: Think deeper, act faster

qwenlm.github.io

Post reply on HN
Print view View original

11–20 of 412 posts

Re: Qwen3: Think deeper, act faster

#11

Post by daemonologist »

It sounds like these models think a lot, seems like the benchmarks are run with a thinking budget of 32k tokens - the full context length. (Paper's not published yet so I'm just going by what's on the website.) Still, hugely impressive if the published benchmarks hold up under real world use - the A3B in particular, outperforming QWQ, could be handy for CPU inference.

Edit: The larger models have 128k context length. 32k thinking comes from the chart which looks like it's for the 235B, so not full length.

Re: Qwen3: Think deeper, act faster

#12

Post by tough »

kadushka wrote: post #4
natrys wrote: post #2

They have got pretty good documentation too[1]. And Looks like we have day 1 support for all major inference stacks, plus so many size choices. Quants are also up because they have already worked with many community quant makers. Not even going into performance, need to test first. But what a stellar release just for attention to all these peripheral details alone. This should be the standard for major release, inste…

they have already worked with many community quant makers I’m curious, who are the community quant makers?

nvm

Re: Qwen3: Think deeper, act faster

#13

Post by minimaxir »

A 0.6B LLM with a 32k context window is interesting, even if it was trained using only distillation (which is not ideal as it misses nuance). That would be a fun base model for fine-tuning.

Out of all the Qwen3 models on Hugging Face, it's the most downloaded/hearted. https://huggingface.co/collections/Qwen/qwen3-67dd247413f0e2...

Re: Qwen3: Think deeper, act faster

#15

Post by natrys »

kadushka wrote: post #4
natrys wrote: post #2

They have got pretty good documentation too[1]. And Looks like we have day 1 support for all major inference stacks, plus so many size choices. Quants are also up because they have already worked with many community quant makers. Not even going into performance, need to test first. But what a stellar release just for attention to all these peripheral details alone. This should be the standard for major release, inste…

they have already worked with many community quant makers I’m curious, who are the community quant makers?

I had Unsloth[1] and Bartowski[2] in mind. Both said on Reddit that Qwen had allowed them access to weights before release to ensure smooth sailing.

[1] https://huggingface.co/unsloth

[2] https://huggingface.co/bartowski

Re: Qwen3: Think deeper, act faster

#16

Post by daemonologist »

sroussey wrote: post #7
natrys wrote: post #2

They have got pretty good documentation too[1]. And Looks like we have day 1 support for all major inference stacks, plus so many size choices. Quants are also up because they have already worked with many community quant makers. Not even going into performance, need to test first. But what a stellar release just for attention to all these peripheral details alone. This should be the standard for major release, inste…

Well, the link to huggingface is broken at the moment.

It's up now: https://huggingface.co/collections/Qwen/qwen3-67dd247413f0e2...

The space loads eventually as well; might just be that HF is under a lot of load.

Re: Qwen3: Think deeper, act faster

#17

Post by cye131 »

These performance numbers look absolutely incredible. The MoE outperforms o1 with 3B active parameters?

We're really getting close to the point where local models are good enough to handle practically every task that most people need to get done.

Re: Qwen3: Think deeper, act faster

#18

Post by jasonjmcghee »

minimaxir wrote: post #13

A 0.6B LLM with a 32k context window is interesting, even if it was trained using only distillation (which is not ideal as it misses nuance). That would be a fun base model for fine-tuning. Out of all the Qwen3 models on Hugging Face, it's the most downloaded/hearted. https://huggingface.co/collections/Qwen/qwen3-67dd247413f0e2...

these 0.5 and 0.6B models etc. are _fantastic_ for using as a draft model in speculative decoding. lm studio makes this super easy to do - i have it on like every model i play with now

my concern on these models though unfortunately is it seems like architectures very a bit so idk how it'll work

Post reply on HN