FRESH

Hacker News

Qwen3-Next

567 points by tosh

by jychang

3 subcomments

Coolest part of Qwen3-Next, in my opinion, (after the linear attention parts) is that they do MTP without adding another un-embedding matrix.
Deepseek R1 also has a MTP layer (layer 61) https://huggingface.co/deepseek-ai/DeepSeek-R1/blob/main/mod...
But Deepseek R1 adds embed_tokens and shared_head.head tensors, which are [129280, 7168] or about 2GB in size at FP8.
Qwen3-Next doesn't have that: https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Instruct/blob...
So it saves a few GB in active parameters for MTP, which is a Big Deal. This is one of the changes that helps significantly speeds up inference.

by Alifatisk

4 subcomments

Alibaba keeps releasing gold content
I just tried Qwen3-Next-80B-A3B on Qwen chat, and it's fast! The quality seem to match Qwen3-235B-A22B. Quite impressive how they achieved this. Can't wait for the benchmarks at Artificial analysis
According to Qwen Chat, Qwen3-Next has the following limits:
Maximum context length: 262,144 tokens
Max summary generation length: 32,768 tokens
This is 2x higher on context length and 4x higher on summary generation compared to Qwen3-235B-A22B, damn
> Qwen3-Next [...] excels in ultra-long-context understanding and complex tasks
Even though their new hybrid architecture is fascinating, I think I'll continue to stick with Qwen2.5-Turbo because it's one of the few models that supports 1M tokens in context length. My use case is uploading large pdfs and ask questions across chapters

by irthomasthomas

5 subcomments

llm -m qwen3-next-80b-a3b-thinking "An ASCII of spongebob"

Here's a classic ASCII art representation of SpongeBob SquarePants:

     .------.
    /  o o  \
   |         |
   |  \___/  |
    \_______/


  llm -m chutes/Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8 \
  "An ASCII of spongebob"

  Here's an ASCII art of SpongeBob SquarePants:

      .--..--..--..--..--..--.
    .' \  (`._   (_)     _   \
  .'    |  '._)         (_)  |
  \ _.')\      .----..--.'  /
  |(_.'  |    /    .-\-.  \---.
  \     0|    |   ( O| O) |    |
   |  _  |  .--.____.'._.-.  |
   \ (_) | o         -` .-`  |
    |    \   |`-._ _ _ _ _\ /
    \    |   |  `. |_||_|   |
     | o |    \_       \    |     -.   .-.
     |.-.  \    `--..-'   O |     `.`-' .'
   _.'  .' |     `-.-'      /-.__   ' .-'
  .' `-.` '.|='=.='=.='=.='=|._/_ `-'.'
  `-._  `.  |________/\_____|    `-.'
     .'   ).| '=' '='\/ '=' |
     `._.`  '---------------'
             //___\   //___\
               ||       ||
               ||_.-.   ||_.-.
              (_.--__) (_.--__)

Meta: I generated a few dozen spongebobs last night on the same model and NONE where as good as this. Most started well but collapsed into decoherence at the end - missing the legs off. Then this morning the very same prompt to the same model API produced a perfect bob on the first attempt. Can utilization affect response quality, if all else remains constant? Or was it just random luck?

Edit: Ok, the very next attempt, a few minutes later, failed, so I guess it is just random, and you have about a 1 in 10 chance of getting a perfect spongebob from qwen3-coder, and ~0 chance with qwen3-next.

by syntaxing

2 subcomments

The craziest part is how far MoE has come thanks to Qwen. This beats all those 72B dense models we’ve had before and runs faster than 14B model depending on how you off load your VRAM and CPU. That’s insane.

by mmmllm

8 subcomments

The same week Oracle is forecasting huge data center demand and the stock is rallying. If these 10x gains in efficiency hold true then this could lead to a lot less demand for Nvidia, Oracle, Coreweave etc

by mynti

1 subcomments

For anyone curious about what the Gated Delta Network is: https://arxiv.org/pdf/2412.06464

by jbellis

2 subcomments

Added Qwen3 Next to the Brokk Power Ranking Open Round (coding benchmark). It's roughly GPT-OSS-20b strength.
Full set of open weight model results: https://brokk.ai/power-ranking?version=openround&models=ds-r...

by lend000

2 subcomments

Prediction: AI will become commoditized ~15 IQ points higher than the state of the art models today, and with larger context, within 4 years as the incremental improvements in training from synthetic data plateaus (we've already used all the "real" data out there) and open source models are cheaply trained on the outputs of the big money models. Then AI development stagnates until someone invents an effective way to use competitive reinforcement learning to train generalized intelligence (similar to how AlphaGo was trained), removing the need for vast quantities of training data. Then, we get real AGI.

by Jgoauh

2 subcomments

Seems impressive, i believe better architectures are really the path forward, i don't think you need more than 100B params taking this model and what GPT OSS 120B can acchieve

by jwr

3 subcomments

Hmm. 80B. These days I am on the lookout for new models in the 32B range, since that is what fits and runs comfortably on my MacBook Pro (M4, 64GB).
I use ollama every day for spam filtering: gemma3:27b works great, but I use gpt-oss:20b on a daily basis because it's so much faster and comparable in performance.

by croemer

0 subcomment

https://archive.is/JH9XL

by KronisLV

0 subcomment

> The Qwen3-Next-80B-A3B-Instruct performs comparably to our flagship model Qwen3-235B-A22B-Instruct-2507, and shows clear advantages in tasks requiring ultra-long context (up to 256K tokens).
This is pretty impressive and a bit like how the GPT-OSS-120B came out and scored pretty well on the benchmarks despite its somewhat limited size.
That said, using LLMs for software dev use cases, I wouldn't call 256K tokens "ultra-long" context, I regularly go over 100K when working on tasks with bigger scope, e.g.:
```
  Look at the existing code related to this functionality and the existing design patterns in the code as well as the guidelines.
  Then plan out the implementation in detail and ask me a few questions along the way to figure the details out better.
  Finally, based on everything so far, do the actual implementation.
  Then look it over and tell me if anything has been missed from the plan, then refactor the code in any number of ways.
```
It could be split up into multiple separate tasks, but I find that the context being more complete (unless the model starts looping garbage, which poisons the context) leads to better results.
My current setup of running Qwen3 Coder 480B on Cerebras bumps into the 131K token limit. If not for the inference speed there (seriously great) and good enough model quality, I'd probably look more in the direction of Gemini or Claude again.

by slimebot80

4 subcomments

Complete newbie here - some questions, if I may!
This stuff can run on a local machine without internet access, correct?
And it can pretty much match Nano Banana? https://github.com/PicoTrex/Awesome-Nano-Banana-images/blob/...
Also -- what are the specs for a machine to run it (even if slowly!)

by davidpfarrell

0 subcomment

Hyped for the release, but bummed they fell for the ‘next’ naming convention.
What will the actual next advanced release be called:
* next-next
* next (2)
* actual-next-final

0 subcomment

by binary132

0 subcomment

I’ve been using gpt-oss-120B with CPU MoE offloading on a 24GB GPU and it’s very usable. Excited to see if I can get good results on this now!

by techsystems

0 subcomment

How does the context length scaling at 256K tokens compare to Llama's 1M in terms of performance? How are the contexts treated differently?

by kristopolous

0 subcomment

I was getting a bunch of strange hallucinations and weird dialog. It sounds like some exasperated person on the verge of a mental breakdown

by jug

1 subcomments

> The Qwen3-Next-80B-A3B-Instruct performs comparably to our flagship model Qwen3-235B-A22B-Instruct-2507
I'm skeptical about these claims. How can this be? Wouldn't there be massive loss of world knowledge? I'm particularly skeptical because a recent trend in Q2 2025 has been benchmaxxing.

by yekanchi

2 subcomments

how much vram it requires?

by esafak

2 subcomments

ICYMI qwen3-max was released last week.

by cchance

0 subcomment

Those rope tests are impressive AF

by pveierland

0 subcomment

> "The content loading failed."
It's amazing how far and how short we've come with software architectures.

by pzo

0 subcomment

would be interesting how they compare to gpt-oss-120b. The latter one runs also very fast and pricing is currently much better than qwen3-next on many providers. Would expect that if this model is such fast pricing should be similar or even lower.

by Western0

2 subcomments

where is gguf?

by boxboxbox4

0 subcomment

[dead]

by pollre

0 subcomment

[dead]

by keyle

0 subcomment

For a model that can run offline, they've nailed how the website can too.
And it appears like it's thinking about it! /s

by croemer

0 subcomment

ERR_NAME_NOT_RESOLVED

by siliconc0w

2 subcomments

All these new datacenters are going to be a huge sunk cost. Why would you pay OpenAI when you can host your own hyper efficient Chinese model for like 90% less cost at 90% of the performance. At that is compared to today's subsidized pricing, which they can't keep up forever.