News for hackers. Stuff that matters.
Re:Qwen3.8 is launching and going open-weight soon by mft_ on Sunday July 19, 2026 @05:26AM (#48966376)
I think everyone is hoping this!

It would be great if they'd release an MoE model somewhere between the 35B size of 3.6 and the 122B version of 3.5 - it could be a great balance of speed and ability for people with reasonably powerful but not insane home computers.

Re:Qwen3.8 is launching and going open-weight soon by embedding-shape on Sunday July 19, 2026 @05:49AM (#48966496)
> the 122B version of 3.5

Yeah, this is what I'm holding out for, the NVFP4 variant of 3.5 122B is blazing fast with reasonable quality and even with max context fits perfectly within 96GB.

Re:Qwen3.8 is launching and going open-weight soon by nsbk on Sunday July 19, 2026 @06:06AM (#48966585)
Sweet. Are you running a Mac Studio Ultra or 4x3090?
Re:Qwen3.8 is launching and going open-weight soon by embedding-shape on Sunday July 19, 2026 @06:14AM (#48966621)
Simpler (and faster): a single RTX Pro 6000 :)
Re:Qwen3.8 is launching and going open-weight soon by androiddrew on Sunday July 19, 2026 @08:00AM (#48967343)
Curious, do you find the 3.5 120B sized MoE works better than the dense 3.6 27B?
Re:Qwen3.8 is launching and going open-weight soon by embedding-shape on Sunday July 19, 2026 @09:02AM (#48967869)
Yeah, Qwen3.5-122B-A10B-NVFP4 produces better responses than Qwen3.6-27B-NVFP4 (both from unsloth), but I'm mostly using them for programming in various ways, mostly Rust, Clojure, Python and JavaScript, and some translations tasks, but not much more than that, so YMMV.

Edit: as a concrete example, I'm working on a "optimization framework via agent harness" right now, Qwen3.6-27B-NVFP4 is often unable to actually complete the optimization within 100 turns, while Qwen3.5-122B-A10B-NVFP4 has no issues finishing within ~50 turns or so.

Re:Qwen3.8 is launching and going open-weight soon by 3abiton on Sunday July 19, 2026 @09:58AM (#48968290)
> Yeah, Qwen3.5-122B-A10B-NVFP4 produces better responses than Qwen3.6-27B-NVFP4 (both from unsloth), but I'm mostly using them for programming in various ways, mostly Rust, Clojure, Python and JavaScript, and some translations tasks, but not much more than that, so YMMV. > > Edit: as a concrete example, I'm working on a "optimization framework via agent harness" right now, Qwen3.6-27B-NVFP4 is often unable to actually complete the optimization within 100 turns, while Qwen3.5-122B-A10B-NVFP4 has no issues finishing within ~50 turns or so.

But why are you using Qwen3.6-27B-NVFP4 compared to the FP8 or full version? In my experience the Q8 of 27B is on par sometimes better than 122B. I am experiemnting witb higher quants for 122B to fit on my Strix Halo, but still, the difference honestly for my workflow is not that much. I just wish they released 3.6-122B version.

Re:Qwen3.8 is launching and going open-weight soon by nicman23 on Monday July 20, 2026 @01:24AM (#48974634)
dense small models do not like quantization. i find 27b fp8 to be smarter albeit less knowledgable versus the 122B
Re:Qwen3.8 is launching and going open-weight soon by nsbk on Sunday July 19, 2026 @06:35AM (#48966714)
Hahaha well played! Hat tip
Re:Qwen3.8 is launching and going open-weight soon by nsbk on Sunday July 19, 2026 @05:46AM (#48966474)
Indeed! That would be the sweet spot for my 2x3090 rig
Re:Qwen3.8 is launching and going open-weight soon by pettijohn on Sunday July 19, 2026 @10:55AM (#48968731)
SO MUCH THIS. I have Strix Halo with 128GB RAM and was a large and fast model like 122B A10B. Here's hoping!
Re:Qwen3.8 is launching and going open-weight soon by cmrdporcupine on Sunday July 19, 2026 @12:04PM (#48969308)
Absolutely. There's a glaring gap in the space for something about the size of Nemotron Super or just under, but actually ... competent.

The fantasy is a 100B or 80B model, but MoE and highly tuned for coding.