Can we get one for how to test websites on a regular sized screen?
That left-hand sidebar needs to be able to scroll on its own, unless you're on a 4k monitor. Otherwise, there's no way to view the bottom half of the list.
I did some work for a sizable org who were pursuing a migration from their in-house auth to cognito. They originally scoped it at two months, and it ended up taking them six to roll it out.
Then they figured out that their cost projection was actually off by an order of magnitude. And also kept getting bitten by peripheral systems newly getting out of sync.
And so then, they embarked upon the journey to roll it all back...
They can have empathy too. Meaning, you can describe a circumstance with many parties involved, with different motivations and conflicting goals -- and it will tell you how these humans would likely feel about each other and the situation at hand.
They can provide text which contains thoughts, reasoning, and empathy, but is that evidence of them having thoughts, reasoning, or empathy?
To paraphrase the Chinese room argument:
Put me in a room with infinite time, paper, pencils, filing cabinets, and give me an English language list of instructions for doing all the neural-net calculations required to run an LLM. Slide your prompt under the door, and, after doing all the calculations, I will slide the next word out, and so I will repeat until the response is done.
In this thought experiment, I'm just following the instructions. So, in this case, what is doing the "thinking" or "reasoning" or "empathy" here?
The inputs going through the model. That act is what is doing the thinking and reasoning and empathizing. Same as when your brain feeds your sensory inputs through its neural network.
Your brain is just a big set of weights/neurons organized into layers that do back and forward propagation, turns out. You are just statistics surrounded by flesh and bones.
A sociopath can also pantomime empathy for their own selfish reasons simply by observing how others behave in similar situations, while feeling nothing special at all.
Hoping this is real. It's too bad to see the signals from Qwen that they will not be releasing a 35B-A3B for the 3.8 lineup. The MoE architecture makes a huge difference for being able to run these local models on reasonable consumer hardware.
Honest question/suggestion for the HN audience- Since Qwen released the weights for Qwen3.8 2.4T-A95B and we already have the staring point of Qwen3.6 35B-A3B, couldn't someone distill the bigger model and make a "pseudo" Qwen3.8 35B-A3B? Sure, it wouldn't be an official Qwen release but couldn't someone improve on Qwen 3.6 and get the thing everyone is asking for?
I am calling this a suggestion for the audience because I don't have the will/resources to do this.
> Base / architecture: Qwen/Qwen3.6-35B-A3B (Qwen3_5MoeForCausalLM, 256 experts, ~3B active). The "3.8" in the name refers to the teacher, not the base.
Not endorsement, haven't run it myself, just found the link.
The smarter 27B is so fast with MTP I've found I really don't need the 35B-A3B. You get around 70tk/s on a M5 Max lowering to around 40tk/s at higher context sizes.
I've benched 3.8 27B being significantly slower and less quality than 3.6 35B-A4B (both 4-bit quant, MTP, both using turboquant 4-bit served by oMLX), to the point that I'm not even using it right now (on an M3 Max). What's your use case and what did you observe? I might be missing something.
I believe you mean 35B-A3B, there was no such thing as A4B. I use 27B and other models for software development, and quite a few research or similar agents. I cannot imagine a world where the old 35B-A3B model is smarter / more capable than 3.8 27B - the difference is night and day for coding at least. Where 35B-A3B was fast and felt like a Haiku model, 27B feels like a strong Sonnet when given the right tools. I don't use turbo quant so can't comment on that, but with the A3B model you're using you probably won't get much from using MTP with small MoE models like that.
People over-quantize things, muck with the temperature and other settings based on superstitions or results from models they think are similar. There's lots of ways to make 3.8 27B dumber.
I've had great success with allowing the model to review optimal settings and my system specs. It comes up with the right configuration. Running Pi harness.
I just had Ornith 1.5 take a moment to configure itself, now it's reviewing a large project I'm working on, so far its really impressive for my needs. Qwen3.8 27b Unsloth(Dynamic 3.0) is also perfect. These two are working together, and I'm in a sweet-spot, I now have all I need.
Yes, A3B, sorry for the typo. I’ve benched it and used it in practice for agentic coding, and I haven’t seen any benefit to it yet. Lightning MTP helps a lot in the perf department.
I would suggest careful benchmarking. I actually tested and benchmarked, and the new Qwen3.8-27B model is actually slower with MTP on my M4 Max. MTP only gains anything when generating long code sequences, which is very unlikely as the model spends most of its time thinking, not generating code, even if you use it for coding (which I don't).
I get ~10 tok/sec on a M1 Ultra, but the real issue is abysmal prompt processing which makes it unusable. On the M1 and M2 series MTP is actively harmful for performance.
27B means it has 27 billion parameters, but how that translates to model size depends on the model architecture. For Qwen 3.5 architecture (same as 3.8) the native size is around 55 GB memory. So you need quantization to fit it in 32 GB.
See the right hand side panel, you see a whole palette of quantizations and their respective sizes. Should give you an idea. Note that these are not the only quantizations available.
Quick tests suggest it’s pretty good at reasoning and tool use (keen to search to check its thinking) and it seems to waste much less time thinking, too.
So it feels very fast.
But it does not seem to be better than Qwen 3.6 35B at coding. A bit worse, I think, though I will test it more.
If you have a machine that can fit a 35B model in VRAM, I would suggest testing Muse Glimmer with (from memory)
Reasoning strength: low
in the system prompt.
Despite being a dense model, this is actually capable of solving code problems faster than the Qwen MoE, despite having only one fifth of the raw token performance.
Personally I find speculative decoding much better strategy than MoE – performance wise it's there at 90-100 t/s on 2x4090, great intelligence – really great fit.
A lot of people, including me, don’t want to bother with GPUs, they’d rather run it on their M1-M5 MacBook. For example the 35B-A3B is very usable even on a M1 64GB MacBook.
Speculative decoding also works on Mac, 64G is more than what I have, m5 max should handle up to ~40 t/s with optimized setup (and with a lot of vram you can get great wins on concurrency – that harness can take advantage of for single user task as well), but agree memory bandwidth in mac or spark is still too slow, next gen for both will be great hardware to have for sure.
MoE is great on systems that lack the VRAM to host the full model. On my 16GB VRAM system, I can get 100 tok/s with Q4 Qwen 3.6 35b a3b, and 15 tok/s with 27b.
MTP is a trade-off, as it pushes some more of the model off the GPU.
I have managed to get usable quants of Laguna S2 and even DeepSeek V4 flash on this setup.
There is clearly some intelligence loss compared to similar sized dense models, but I feel like it stomps on the 9-12b models I could run fully on GPU
> The MoE architecture makes a huge difference for being able to run these local models on reasonable consumer hardware
That's not true. For computers without unified memory architecture (which is the vast majority) VRAM capacity is the bottleneck for local models. In that case a dense model can deliver significantly more intelligence than an MoE model of the same size. And for a typical consumer/gamer Nvidia GPU, dense models are fast enough.
Expert offloading significantly helps with the VRAM capacity.
Most MoE architectures have a few experts that are always running; this, the router, KV, and whatever else you have space for can stay in fast VRAM; and the remaining experts can be offloaded.
You can do so many different things with a computer. If you spend hours watching AI generated YouTube shorts instead of playing outside with your kids, then yeah, that's too bad.
But if you spend all those hours building a retro/modern sim city game pair programming with your 12yo, designing the artwork, crafting the story arc, writing the music, finding and fixing bugs, building new levels -- you don't need to regret that you weren't idling around at the park all that time.
I am pro AI, and also deeply share this sentiment.
I think sometimes people have a big long thread with Claude where they feel enthusiastic about the back and forth being very productive -- and then they genuinely want to share the summary of the conversation with their colleagues so that they can feel it too. It's not just typing a short prompt and copy/pasting the response.
But, Claude, especially Opus is so unnecessarily flowery, and smug -- and it's so identifiably characteristic.
It somehow rubs salt in the wound: "I didn't take the time to write this out in my own words, so now you get to listen to this petulant asshole mansplain it to you".
If it would get to the point, and with some humility, it would be a much different proposition.
That left-hand sidebar needs to be able to scroll on its own, unless you're on a 4k monitor. Otherwise, there's no way to view the bottom half of the list.
reply