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.
parsimo2010 16 hours ago [-]
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.
WASDx 15 hours ago [-]
"Qwen3.8 35B-A3B" and 4B/9B variants are already on huggingface distilled by hobbyists.
karlkloss 18 minutes ago [-]
>I can't find any. Do you have a link?
halJordan 14 hours ago [-]
Yes, of course. But no one really wants to be the guy actually renting an entire B300.
boznz 14 hours ago [-]
..And there lies the problem.
smcleod 13 hours ago [-]
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.
jwr 1 hours ago [-]
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'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.
smcleod 13 hours ago [-]
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.
xscott 12 hours ago [-]
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.
razster 4 hours ago [-]
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.
seanmcdirmid 11 hours ago [-]
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.
quinncom 13 hours ago [-]
I only get ~4 tok/sec on a M1 Pro with MTP.
regexorcist 7 hours ago [-]
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.
meatmanek 10 hours ago [-]
What quant, what runtime?
smcleod 9 hours ago [-]
AWQ 5bit, oQ5. oMLX.
Muromec 13 hours ago [-]
Does 27b mean it fits one 32GB GPU?
kzrdude 2 hours ago [-]
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.
smcleod 13 hours ago [-]
It doesn't mean that, but yes it would (with a 5 bit quant).
dofm 15 hours ago [-]
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.
mirekrusin 15 hours ago [-]
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.
d4rkp4ttern 15 hours ago [-]
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.
mirekrusin 14 hours ago [-]
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.
smcleod 13 hours ago [-]
I get around 70tk/s on the m5 max, with 5bit AWQ / oQ5 slowing only to around 40tk/s at higher context.
c0m47053 14 hours ago [-]
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
Sha1rholder 10 hours ago [-]
> 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.
dannyw 2 hours ago [-]
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.
slim 9 hours ago [-]
llama.cpp can run MoE with some layers in vram and some layers in ram
huseyinkeles 11 hours ago [-]
I've been experimenting with it on a M4 Pro 24G for the last few hours and it's been very promising using 32k context. getting around 30-40 tps
With Qwen3.8 27B I could not get anywhere near 32k context window, that made it very unusable for agentic coding, although it was very smart.
regexorcist 7 hours ago [-]
What signals? Looking more like a tiered release.
hgoel 12 hours ago [-]
The way it was said made it sound like they had something better than a 35B-A3B coming.
vkaku 11 hours ago [-]
This is all real. More real things coming soon.
verdverm 15 hours ago [-]
I'm running qwen3.8 27B dense on reasonable hardware (oem spark)
tbh, I have stopped using MoE in the name of speed, the dense (with more active parameters) makes a real difference in output quality
What did you use it for? I've found it somewhat capable but not worth to actually use it (1 9B, that is)
jonesy827 16 hours ago [-]
I've been using the 35B-A3B today for some web scraping work, and it has been on par with Qwen3.8 27B at a much higher speed and at a higher quant (q4 vs q8). I'm impressed.
jakswa 16 hours ago [-]
I had to go down to UD-Q3_K_XL for Qwen 3.8 27B to get it to fit in VRAM and be usable, but I worry I'm gutting its intelligence somewhat. I too am interested in faster + more-usable alternative that can exchange blows with the Q3-dumbed 27B.
jadbox 16 hours ago [-]
I need someone to run actual benchmarks between the two.
swatcoder 16 hours ago [-]
Benchmarks are the BMI of model evaluation.
They may have utility in trying to look at the whole landscape of models, but are very misleading when it comes to making 1:1 comparisons or in developing confidence at to how a given model will deliver on your workflow.
NitpickLawyer 15 hours ago [-]
Only relevant benchmarks are those you make yourself, targeted specifically for your workflows. Anything else is just number go up on a pretty graph, and every model out there is probably benchmaxxed to hell on the public ones anyway. Keep yours private.
gertlabs 15 hours ago [-]
These models have gotten a fair amount of attention -- we're hoping it's enough to get them added to some reliable inference providers and OpenRouter, at which point we'll run them on our full benchmark suite.
bigcat12345678 17 hours ago [-]
How is ornith-1.5's base model developed? Is the base model one of the Open weights models, or one pre trained by ornith team from scratch? I couldn't find information to answer this question in the article.
goldemerald 17 hours ago [-]
It looks like they post-trained Qwen3.6. Interesting to see how far they could improve it with they harness/algorithm.
lsb 12 hours ago [-]
The page has comparisons with Qwen 3.6 27b and I’d love to see comparisons with Qwen 3.8 27b, the newer one is much more capable!
ricardobeat 11 hours ago [-]
It's somewhat close, but a lot worse at code it seems. Qwen 3.8 is a wild improvement over 3.6.
Interestingly, in my own benchmark and testing (in the hopes of finding a good-enough local model to run a personal assistant agent), Ornith-1.0-9B was worse than Qwen3.5-9B which according to their scores should've been reversed.
I will definitely pass Ornith-1.5-9B through the gauntlet as well!
17 hours ago [-]
colingauvin 16 hours ago [-]
397 is just too big for two Sparks even at NVFP4. Wish they had made this just a tiny bit smaller.
kees99 15 hours ago [-]
Ornith-1.5-397B is derived from Qwen3.5-397B-A17B via post-training. That process preserves exact parameter count.
ggcr 15 hours ago [-]
They should've included Qwen3.5-397B-A17B in the benches then :/
colingauvin 15 hours ago [-]
Ah thanks for the background. Well, time to learn how to quant things down!
rbanffy 12 hours ago [-]
It’s time for me to upgrade the main server in my home lab and I’m thinking about which machine should I have.
What kind of hardware you’d need to run the 397B one at an acceptable speed?
colingauvin 12 hours ago [-]
3 DGX Sparks with 400 Gb interconnects in a loop.
rbanffy 2 hours ago [-]
I was looking for something more general-purpose, such as a decommissioned quad-socket x86 with 80 or more AVX512 cores and a terabyte of RAM, which I suspect will be a strong limiting factor. The possibility of adding some older Nvidia accelerators is interesting as well even though they won’t be able to hold the whole model in HBM and will need to stream it over PCIe 4, which sucks.
That way, when not running inference, the machine can easily host VMs for other experiments and general housekeeping functions.
I wonder what their angle is going to be; the scene is crowded, and they don't do serving.
garo-pro 15 hours ago [-]
Across five cases it reliably claims to be Claude without being able to name a specific version.
wgd 15 hours ago [-]
Nobody (with the probable exception of Anthropic given their work on character training) really trains models on their identity and Claude is the only AI persona that's well-defined so if you put yourself into the AI's shoes it's a pretty reasonable guess that it might be Claude. I've had basically every open model claim it's Claude when the topic comes up.
orbital-decay 14 hours ago [-]
Anthropic changed their character training policy many times, and the name is probably separate from it anyway (besides the bits from the constitution etc). Name training usually comes last in all models, if at all, and it's pretty shallow. Certain Claude models say they are Qwen or Deepseek when asked in Chinese, for example.
wgd 12 hours ago [-]
Yeah, Claude is actually surprisingly unsure of his identity considering that their most recent publication on their constitutional AI training literally had graphs demonstrating how certain properties differed based on whether they were phrased as questions about "Claude" versus "You", but it's actually that which makes me fairly confident that they're probably doing _something_ to try and close that gap.
More generally, their current approach to constitutional AI pretty much only makes sense if they believe that they can first teach the model what the Claude character is like and also teach the model that the persona responding is Claude, so I figure that has to be part of the pipeline even if they're not very good at it.
I'll be comparing the 9B vs Ling 3 Tiny (8B-A1B) as a scout model. Ling tiny is so fast but can be a little too dumb. Hope the 9B strikes a good middleground even if dense/slower.
AIorNot 10 hours ago [-]
Can someone parse all that AI generated blather in the post and tell me clearly:
1. Is this self improvement at the model level (updated weights or memory, KV etc) or just by adding agentic code harnesess to guide the output better?
Thank you
kzrdude 2 hours ago [-]
The former
tangjurine 17 hours ago [-]
This looks cool
wyre 16 hours ago [-]
This is exciting.
Their 9B model benchmarks competitively with Sonnet 4 which is pretty cool to have such a small model compared to one that came out 10 months ago.
I’m curious how providers will price their 397B model.
htrp 15 hours ago [-]
Another day another startup claiming some vague version of RSI to try to close their round.
Rendered at 09:43:17 GMT+0000 (Coordinated Universal Time) with Vercel.
I am calling this a suggestion for the audience because I don't have the will/resources to do this.
I get 20 tokens/s on an M4 Max (larger GPU).
Go here: https://huggingface.co/unsloth/Qwen3.8-27B-GGUF
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.
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)
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.
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
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.
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.
With Qwen3.8 27B I could not get anywhere near 32k context window, that made it very unusable for agentic coding, although it was very smart.
tbh, I have stopped using MoE in the name of speed, the dense (with more active parameters) makes a real difference in output quality
They may have utility in trying to look at the whole landscape of models, but are very misleading when it comes to making 1:1 comparisons or in developing confidence at to how a given model will deliver on your workflow.
I will definitely pass Ornith-1.5-9B through the gauntlet as well!
What kind of hardware you’d need to run the 397B one at an acceptable speed?
That way, when not running inference, the machine can easily host VMs for other experiments and general housekeeping functions.
I wonder what their angle is going to be; the scene is crowded, and they don't do serving.
More generally, their current approach to constitutional AI pretty much only makes sense if they believe that they can first teach the model what the Claude character is like and also teach the model that the persona responding is Claude, so I figure that has to be part of the pipeline even if they're not very good at it.
1. Is this self improvement at the model level (updated weights or memory, KV etc) or just by adding agentic code harnesess to guide the output better?
Thank you
Their 9B model benchmarks competitively with Sonnet 4 which is pretty cool to have such a small model compared to one that came out 10 months ago.
I’m curious how providers will price their 397B model.