I'm a huge open model fan, and have used them since forever, even have daily drivers for on-prem dev, but no. They do not beat opus on real-world usage.
Qwen models are impressively good for what they are, are "good enough" for plenty tasks, can be ran locally on decently priced hardware, and so on. They certainly have their uses, and the field in general has advanced faster than my early expectations. But to compare a 27B model to SotA behemoths from a few months ago is doing everyone a disservice, especially people who pick it up, try to use them just like API models, and leave disappointed and confused. Number goes up on a benchmark isn't it.
We have an internal eval that measures performance on tasks for a handful of embedded systems repos for our mmWave radios (mostly Rust, some C for microcontroller stuff). Qwen3.6-27B scores only 4% lower for pass@1, n=250 compared to Opus-4.8.
For the labeled dataset, the average PR size they're being measured against is around 1.5k SLOC.
This is very much "real-world usage" for us. The sort of change sets that come in daily/weekly and are solving non-trivial issues in the respective codebases.
As is usually the case, the most broad claims from both the labs and from the consequent pushback are talking past each other.
> ...but no. They do not beat opus on real-world usage.
I agree, but then we just need meaningful benchmarks that clearly show that! Otherwise it's hand waving about something that should be put on paper in quantifiable terms.
> but then we just need meaningful benchmarks that clearly show that!
That's the rub. AI benchmarks are IMO, by and large totally unreliable. We think of them as similar to traditional benchmarks of deterministic processes where the number of variables is low. But they're anything but that. Non-deterministic processes with an astounding number of variables and fuzzy acceptance criteria.
It leads to results like these, where if you take it at face value, the only conclusion you can draw is "wow Anthropic must be stupid if Opus takes 1T parameters to do what Qwen can do in 27B."
I hope really badly that we'll get a new 35B A3B or similar MoE model!
I also miss the Qwen 3 Coder Next, which was 80B A3B, there are quite a few use cases where a non-dense model <100B would be the sweet spot (when you have the VRAM but not the TDP or compute power). Heck, I'd gladly take A5B or A8B or even A10B as a sort of middle ground.
Same here! Qwen3.6-35B-A3B is the only local model I've found that runs reasonably on my iGPU. Looks like me and and my noisily-wheezing laptop will be sitting out this upgrade.
Model benchmarks are useful, to a point, but it is the long tail of things you do with the model that determines if it's good at a wide range of activities. Ant/OAI, to their credit, build their models -- even the small ones -- so they follow instructions and do tool calling well, without the system prompts confusing them. This is especially important for long-horizon tool calling.
So one open weight model might "meet" Opus or whatever on benchmarks, but then fail to follow a simple answer format and also tool call correctly. The models are whipped to within an inch of their lives to strictly adhere to their post training quality gates.
I hope the bonsai team makes another 1bit quant of this model (or releases code/instructions on how to do it), using the Qwen3.6 27B on my 16GB mac mini has been wild . The 1bit quant feels like opus level… for the first couple turns. Then it has trouble eg switching from plan mode to act mode. This is mostly mitigated by starting a new session. (tbf this limitation is called out on the hf page)
I saw unsloth has 1bit quants too so I might check that out, anybody have experience with those?
If the benchmarks are a real indication, we now have a local model that is runnable on a high-end personal PC that trades blows with the leading model Claude Opus 4.6 Max from half a year ago.
can you tell me ideas of usecases of 9 or 10B language models ? I cant find any usecases other than training a lora on them to give good bash commands for example
Can anyone who has that specific personal test he tries on different models , and tries this model , to tell us here if possible , how good or bad is this new model ? compared to others ?
There is a down to earth guy on YT that performs a series of tests against LLMs running on non-god-tier commodity hardware. He will likely be testing this soon enough.
Any tips on the best approach at running this at an M4 Max 128GB? Token throughput was a bit slow with the last 27B one (MLX), ended up using the A3B variant but if I could get this one to reasonable speed I'd much prefer it.
Go for a slightly more quantised version, and experiment with different MTP settings. I find that MLX versions are marginally faster on my 64GB M1 Max, but I usually use Unsloth's GGUFs via llama.cpp as there's a much greater range of quants available and I prefer llama.cpp. MTP sometimes also helps a little, but I suspect it's less helpful on my system than others.
Unfortunately, that chip just doesn't really have the memory bandwidth to run this (or nearly any) model at acceptable speeds. I have the exact same chip (M4 Max 128GB) and I've been trying to optimize a completely purpose-built implementation with Fable and it's just not possible. Even if you could reach the full 576GB/s, it's just physically impossible to exceed these numbers with the model's architecture:
2 bpw - ~85.7t/s
3 bpw - ~58.0t/s
4 bpw - ~43.9t/s
6 bpw - ~29.5t/s
8 bpw - ~22.2t/s
16 bpw - ~11.2t/s
without cheating. You'd have to exclude layers, skip operations, etc. basically do stuff the model wasn't trained for. And speed collapses so fast with context that even 2 bpw would be looking at ~37.6t/s after just 128K tokens.
MTP only improves the situation by up to 2x in the ideal case, while drastically reducing the performance floor. While optimizing a 9B model on this hardware, I've found that the GPU just doesn't have enough FLOPS to handle speculating more than one or two tokens ahead on a single stream, regardless of quant level, simply because of the arithmetic cost of the forward pass. The 27B model would be even more expensive than that.
I wouldn't get my hopes up for the 35B-A3B either. Not only is it reportedly much less intelligent, but I hit the exact same 85t/s wall in practice (again with highly specialized inference).
Without speculation I can reach around 120t/s on Qwen3.5-9B and with n-gram speculation (not even MTP; this derivative didn't come with one) around about 150t/s on average. This is on the very very edge of what I'd consider acceptable for me to even consider using such a compact model. YMMV due to the silicon lottery but the situation isn't good.
Beats Opus 4.7 Max (w/ Claude Code) on DeepSWE (42.2 vs 40). Looks like Qwen's 27B models continue to pack some punch.
Unsloth's GGUF quants are up: https://huggingface.co/unsloth/Qwen3.8-27B-GGUF
> Beats Opus 4.7 Max
I'm a huge open model fan, and have used them since forever, even have daily drivers for on-prem dev, but no. They do not beat opus on real-world usage.
Qwen models are impressively good for what they are, are "good enough" for plenty tasks, can be ran locally on decently priced hardware, and so on. They certainly have their uses, and the field in general has advanced faster than my early expectations. But to compare a 27B model to SotA behemoths from a few months ago is doing everyone a disservice, especially people who pick it up, try to use them just like API models, and leave disappointed and confused. Number goes up on a benchmark isn't it.
> They do not beat opus on real-world usage
We have an internal eval that measures performance on tasks for a handful of embedded systems repos for our mmWave radios (mostly Rust, some C for microcontroller stuff). Qwen3.6-27B scores only 4% lower for pass@1, n=250 compared to Opus-4.8.
For the labeled dataset, the average PR size they're being measured against is around 1.5k SLOC.
This is very much "real-world usage" for us. The sort of change sets that come in daily/weekly and are solving non-trivial issues in the respective codebases.
As is usually the case, the most broad claims from both the labs and from the consequent pushback are talking past each other.
> ...but no. They do not beat opus on real-world usage.
I agree, but then we just need meaningful benchmarks that clearly show that! Otherwise it's hand waving about something that should be put on paper in quantifiable terms.
A wise man once said, "not everything that counts can be counted, and not everything that can be counted counts".
> but then we just need meaningful benchmarks that clearly show that!
That's the rub. AI benchmarks are IMO, by and large totally unreliable. We think of them as similar to traditional benchmarks of deterministic processes where the number of variables is low. But they're anything but that. Non-deterministic processes with an astounding number of variables and fuzzy acceptance criteria.
It leads to results like these, where if you take it at face value, the only conclusion you can draw is "wow Anthropic must be stupid if Opus takes 1T parameters to do what Qwen can do in 27B."
Only useful benchmarks are those you (in particular) don't have access to.
Is there any advantage to using the model from Unsloth compared with https://huggingface.co/Qwen/Qwen3.8-27B-FP8 ?
Depends on what software/hardware you'll run it. GGUFs from Unsloth can run on pretty much every single potato; full weights need beefy gpus
Unsloth one is gguf for llama.cpp (and some other on-device engines).
So advantage is not having to produce your own quantisation / gguf from .safetensors you've linked.
Run the unsloth if you are using llama.cpp (GGUF)
Run the one you linked if you are running vllm (safetensors)
I wish each quant was benchmarked on the same tests as the original network so we could compare their performance
That's crazy, considering the massive size difference. But the small Qwen models are known for punching above their weight.
I hope really badly that we'll get a new 35B A3B or similar MoE model!
I also miss the Qwen 3 Coder Next, which was 80B A3B, there are quite a few use cases where a non-dense model <100B would be the sweet spot (when you have the VRAM but not the TDP or compute power). Heck, I'd gladly take A5B or A8B or even A10B as a sort of middle ground.
Also alternate link for viewing the images without signing in: https://xcancel.com/Alibaba_Qwen/status/2088280182356611304
> I'd gladly take A5B or A8B or even A10B as a sort of middle ground.
Whats up with focusing on the active param count? Do yall fiddle with the weights or something?
Same here! Qwen3.6-35B-A3B is the only local model I've found that runs reasonably on my iGPU. Looks like me and and my noisily-wheezing laptop will be sitting out this upgrade.
I'm hoping too that they'll put out some MoE variants.
Qwen3.5:122b:a10b can run about twice as fast as this 27b dense model.
I'm wondering whether any provider can offer this for cheaper $/token than the new DSv4 Flash, which is both cheaper and smarter :/
Completely local use is a different story, of course.
https://xcancel.com/Alibaba_Qwen/status/2088280182356611304
I could kiss you right now
better than opus 4.6 max ╰(°□°)╯
Model benchmarks are useful, to a point, but it is the long tail of things you do with the model that determines if it's good at a wide range of activities. Ant/OAI, to their credit, build their models -- even the small ones -- so they follow instructions and do tool calling well, without the system prompts confusing them. This is especially important for long-horizon tool calling.
So one open weight model might "meet" Opus or whatever on benchmarks, but then fail to follow a simple answer format and also tool call correctly. The models are whipped to within an inch of their lives to strictly adhere to their post training quality gates.
I hope the bonsai team makes another 1bit quant of this model (or releases code/instructions on how to do it), using the Qwen3.6 27B on my 16GB mac mini has been wild . The 1bit quant feels like opus level… for the first couple turns. Then it has trouble eg switching from plan mode to act mode. This is mostly mitigated by starting a new session. (tbf this limitation is called out on the hf page)
I saw unsloth has 1bit quants too so I might check that out, anybody have experience with those?
If the benchmarks are a real indication, we now have a local model that is runnable on a high-end personal PC that trades blows with the leading model Claude Opus 4.6 Max from half a year ago.
Insane if that is the case. Downloading now!
27b dense model at Opus 4.6 level
Opus at home
I hope there also will be a new ~10b variant
can you tell me ideas of usecases of 9 or 10B language models ? I cant find any usecases other than training a lora on them to give good bash commands for example
Can anyone who has that specific personal test he tries on different models , and tries this model , to tell us here if possible , how good or bad is this new model ? compared to others ?
I only trust those users genuine personal tests
There is a down to earth guy on YT that performs a series of tests against LLMs running on non-god-tier commodity hardware. He will likely be testing this soon enough.
https://www.youtube.com/@lukesdevlab
I don't know if that is what you are looking for or not and as always your experiences may be different.
Any tips on the best approach at running this at an M4 Max 128GB? Token throughput was a bit slow with the last 27B one (MLX), ended up using the A3B variant but if I could get this one to reasonable speed I'd much prefer it.
Go for a slightly more quantised version, and experiment with different MTP settings. I find that MLX versions are marginally faster on my 64GB M1 Max, but I usually use Unsloth's GGUFs via llama.cpp as there's a much greater range of quants available and I prefer llama.cpp. MTP sometimes also helps a little, but I suspect it's less helpful on my system than others.
Unsloth: https://huggingface.co/unsloth/Qwen3.8-27B-GGUF
This might work for you, but I didn't get on very well with MTPLX when I tried it a while back; YMMV: https://huggingface.co/Youssofal/Qwen3.8-27B-MTPLX-Optimized...
Unfortunately, that chip just doesn't really have the memory bandwidth to run this (or nearly any) model at acceptable speeds. I have the exact same chip (M4 Max 128GB) and I've been trying to optimize a completely purpose-built implementation with Fable and it's just not possible. Even if you could reach the full 576GB/s, it's just physically impossible to exceed these numbers with the model's architecture:
2 bpw - ~85.7t/s
3 bpw - ~58.0t/s
4 bpw - ~43.9t/s
6 bpw - ~29.5t/s
8 bpw - ~22.2t/s
16 bpw - ~11.2t/s
without cheating. You'd have to exclude layers, skip operations, etc. basically do stuff the model wasn't trained for. And speed collapses so fast with context that even 2 bpw would be looking at ~37.6t/s after just 128K tokens.
MTP only improves the situation by up to 2x in the ideal case, while drastically reducing the performance floor. While optimizing a 9B model on this hardware, I've found that the GPU just doesn't have enough FLOPS to handle speculating more than one or two tokens ahead on a single stream, regardless of quant level, simply because of the arithmetic cost of the forward pass. The 27B model would be even more expensive than that.
I wouldn't get my hopes up for the 35B-A3B either. Not only is it reportedly much less intelligent, but I hit the exact same 85t/s wall in practice (again with highly specialized inference).
Without speculation I can reach around 120t/s on Qwen3.5-9B and with n-gram speculation (not even MTP; this derivative didn't come with one) around about 150t/s on average. This is on the very very edge of what I'd consider acceptable for me to even consider using such a compact model. YMMV due to the silicon lottery but the situation isn't good.
Check out MTPLX and limit your context size.
These are massive improvements - and something you can actually run on a laptop.
Looks like a pretty significant improvement on the DeepSWE benchmark compared to the previous 27B model.
q4km is about 48 tps on a 4090. my llama.cpp params are --flash-attn on --parallel 1 --load-mode mmap
With spec decode should easily get to >100tps
on my dual 3090s qwen 3.5 27b was running at around 110tps using the config from https://github.com/noonghunna/club-3090
make that 200tps on a single 5090, 4x faster than opus https://x.com/radixark/status/2088285681131110446
devs about to get handed a two 5090 box each and told to max that out
3.6-27B on little-coder was already mind blowing. looking forward to this guy!
Seems to be SOTA for its size. Hopefully independent benchmarks will come soon.
Monstrous benchmarks! Hoping it is not benchmaxxed.
Welcome deepseek flash flash!
NICE, i've been waiting for this drop, thanks for posting this
This with ddg mcp to fill in world knowledge. Are local models the future when computer architectures catch up?
remember to let llama.cpp catch up to anything new in this model. Save your judgment until about 2 weeks of use.
'Good' news, there seems to be nothing new architecture-wise. Same as Qwen 3.5 and 3.6.
My Strix Halo is about to go overdrive!
also cool: Qwen 3.8 27b is multi modal!
another thread: https://news.ycombinator.com/item?id=49294502
need another fable uncensored merge with 3.8, really curious what it can deliver