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

Qwen3.7-Max: The Agent Frontier

qwen.ai

Post reply on HN
Print view View original

251–260 of 317 posts

Re: Qwen3.7-Max: The Agent Frontier

#251

Post by rao-v »

DiabloD3 wrote: post #227
julianlam wrote: post #178

Earlier quoted context omitted.

May I ask why the M instead of XL? Obviously bigger != better but I don't know what the differences are.

These are dynamic quants, and they're basically just an indication of how far away from the desired quant it is allowed to go to achieve the goal. Generally, unsloth's toolchain moves quants up, rarely down. * _0 and _1 do not use K quant and scales 32x32 blocks according to the original (B)F16 values; _0 scales the block using the original max and min values. _1 does this per row instead of per block. * K quants do…

Hey some of us are on hardware (gfx906 based Radeon MI50s with 32GB of stupidly fast VRAM and basically no compute) that inference significantly faster with Q_0 and Q_1 quants

Re: Qwen3.7-Max: The Agent Frontier

#252

Post by chr15m »

briga wrote: post #109

I was getting dangerously close to my weekly Claude Code limit last night so I had Claude set up Qwen3.6 with llama.cpp and OpenCode. Honestly it's a great (free!) alternative to Claude Code--certainly more than good enough for a lot of smaller less complex tasks. I'm excited to try this new version. The fact that open-source models are so close to the frontier is very impressive.

This new version is not something you'll be able to run locally. It's a "cloud" model and likely too beefy if they do release the weights.

Re: Qwen3.7-Max: The Agent Frontier

#253

Post by nomel »

jorvi wrote: post #237
speed_spread wrote: post #215

Earlier quoted context omitted.

Yes. A model that can answer "I don't know" would be much more trustable than the current used car salesman we have now.

Its very annoying this has been in the capability of models since the very beginning. It could check how probable its token values are and if those fall below a certain threshold either say "I don't know", or output the most probable (well, more like least improbable) tokens but give a very clear, very strong warning that it is a shot in the dark and likely to contain hallucinations. But no, Google and OpenAI would r…

Yeah, I never understood why the top n statistics weren't included in the chat interfaces, to color the text!

Re: Qwen3.7-Max: The Agent Frontier

#254

Post by tokenscoper »

jorvi wrote: post #237
speed_spread wrote: post #215

Earlier quoted context omitted.

Yes. A model that can answer "I don't know" would be much more trustable than the current used car salesman we have now.

Its very annoying this has been in the capability of models since the very beginning. It could check how probable its token values are and if those fall below a certain threshold either say "I don't know", or output the most probable (well, more like least improbable) tokens but give a very clear, very strong warning that it is a shot in the dark and likely to contain hallucinations. But no, Google and OpenAI would r…

I don't have much to add other than this observation that we seem to have moved away from eating one small rock per day for nutritional value, and adding gasoline in spaghetti.

The glue on pizza reference brought back memories :)

Re: Qwen3.7-Max: The Agent Frontier

#255

Post by rayboy1995 »

satvikpendem wrote: post #199
rayboy1995 wrote: post #161

Earlier quoted context omitted.

Thanks!! I had disabled that previously while debugging, I can confirm this is helping accuracy from what I can tell so far. (And speed since the cache is preserved more often!)

Use the MTP models which 2x token generation speed, for example: https://unsloth.ai/docs/models/qwen3.6#mtp-guide

Very interesting I'll have to check this out thank you. This is why I love HN.

Re: Qwen3.7-Max: The Agent Frontier

#256

Post by grumple »

When I click on the link to Alibaba Cloud Model Studio from the linked post, that page sends my CPU (9950X3D) to 100%. Which is just... impressive. Is this a js based crypto miner? Or some strange browser based particle display? Super weird.

Re: Qwen3.7-Max: The Agent Frontier

#257

Post by nl »

girvo wrote: post #242
goldenarm wrote: post #61

The non-hallucination rate in AA-omniscience is SOTA, better than Opus 4.7, Gemini 3.1 Pro and GPT5.5! Congrats to the team

The big question for me having used a lot of these SOTA chinese models is: what is its token efficiency like? Running Step 3.5 Flash locally for example, it's an amazingly capable model all things considered, but it's token efficiency is so bad that it gets out performed by most others wall-clock time (even with my MTP-support for it hacked in to llama.cpp: despite being trained on three heads, MTP 2 is the sweet spo…

Yes it's a big thing that people are slowly becoming more aware of.

Nvidia models are even worse than Qwen! https://sql-benchmark.nicklothian.com/#token-efficiency-and-... (mouse over the cells for token counts and click for traces)

Gemma 4 is good for this, as AA notes:

> Gemma 4 31B is notably token efficient, using 39M output tokens to run the Intelligence Index vs 98M for Qwen3.5 27B (Reasoning). This is ~2.5x fewer output tokens for a model scoring 3 points lower. For context, the other models at the 42-point intelligence level also use significantly more tokens: MiniMax-M2.5 (56M), DeepSeek V3.2 (Reasoning, 61M), and GLM-4.7 (Reasoning, 167M)

https://artificialanalysis.ai/articles/gemma-4-everything-yo...

Re: Qwen3.7-Max: The Agent Frontier

#258

Post by miki123211 »

jorvi wrote: post #237
speed_spread wrote: post #215

Earlier quoted context omitted.

Yes. A model that can answer "I don't know" would be much more trustable than the current used car salesman we have now.

Its very annoying this has been in the capability of models since the very beginning. It could check how probable its token values are and if those fall below a certain threshold either say "I don't know", or output the most probable (well, more like least improbable) tokens but give a very clear, very strong warning that it is a shot in the dark and likely to contain hallucinations. But no, Google and OpenAI would r…

It can't, because top n isn't always reliable.

Hallucination detection is an open problem. If it were that simple, people would indeed "just" do it.

Basically the problem is that LLMs aren't trained on things they don't know; an alternative way of saying this is that they're not trained on things they're not trained on, which is obviously true.

When you RL a model and it answers incorrectly, you don't teach it to answer "I don't know", you teach it to answer correctly instead. This makes it very hard for it to realize when it doesn't know things.

Re: Qwen3.7-Max: The Agent Frontier

#259

Post by aicantdeny »

jampekka wrote: post #193
throawayonthe wrote: post #81

Earlier quoted context omitted.

referencing this: https://artificialanalysis.ai/evaluations/omniscience?models... (had to add it to the chart, wasn't displayed by default. is it the lowest rate in the datasetor no?)

This counts only incorrect answers though. A model can get 0% hallucination rate just by refusing to answer all questions.

> by refusing to answer all questions.

Cool, precisely the thing other AI is too stupid to do when they don't have the necessary knowledge.

Re: Qwen3.7-Max: The Agent Frontier

#260

Post by plufz »

gcr wrote: post #153
plufz wrote: post #113

Earlier quoted context omitted.

Which exact model are you using? And with which parameters and quant? And on what hardware? Are you using any specific MCPs or other tools to optimize performance like context-mode or dynamic context pruning? I’ve used local models a reasonable amount before but I’m just starting out with opencode. Haven’t had great results yet but really want this to work for simpler tasks. My opencode newly installed is also having…

here's a simple setup to get you started on an Apple M1 Max from 2021 with 32GB VRAM. it will download 20GB of models to `~/.cache/huggingface/hub`, which you can delete when you're done. /Users/gcr/llama.cpp/build/bin/llama-server -hf unsloth/Qwen3.6-35B-A3B-GGUF:Q4_K_M --no-mmproj-offload --fit on -c 65536 # edit to taste --reasoning on --chat-template-kwargs '{"preserve_thinking": true}' --sleep-idle-seconds 90 #…

Thanks a million!
Post reply on HN