I wish that all announcements of models would show what (consumer) hardware you can run this on today, costs and tok/s.
Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
11–20 of 482 posts
Re: Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
#12This is getting very close to fit a single 3090 with 24gb VRAM :)
Re: Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
#13Has anyone tested it at home yet and wants to share early impressions?
I have been kicking the tires for about 40 minutes since it downloaded and it seems excellent at general tasks, image comprehension and coding/tool-calling (using VLLM to serve it). I think it squeaks past Gemma4 but it's hard to tell yet.
good to hear! Do you mind sharing your setup and tokens / seconds performance ?
Re: Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
#14Good news!
Friendly reminder: wait a couple weeks to judge the ”final” quality of these free models. Many of them suffer from hidden bugs when connected to an inference backend or bad configs that slow them down. The dev community usually takes a week or two to find the most glaring issues. Some of them may require patches to tools like llama.cpp, and some require users to avoid specific default options.
Gemma 4 had some issues that were ironed out within a week or two. This model is likely no different. Take initial impressions with a grain of salt.
Re: Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
#15Unsloth quants available:
Re: Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
#16I wish that all announcements of models would show what (consumer) hardware you can run this on today, costs and tok/s.
depends on format, compute type, quantization and kv cache size.
Re: Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
#17A bit skeptical about a 27B model comparable to opus...
Some of these benchmarks are supposedly easy to game. Which ones should we pay attention to?
well, your own, unleaked ones, representing your real workloads.
if you can't afford to do that, look at a lot of them, eg. on artificialanalysis.com they merge multiple benchmarks across weighted categories and build an Intelligence Score, Coding Score and Agentic score.
Re: Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
#18Re: Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
#19This is getting very close to fit a single 3090 with 24gb VRAM :)
Yup! Smaller quants will fit within 24GB but they might sacrifice context length.
I’m excited to try out the MLX version to see if 32GB of memory from a Pro M-series Mac can get some acceptable tok/s with longer context. HuggingFace has uploaded some MLX versions already.
Re: Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model
#20This is getting very close to fit a single 3090 with 24gb VRAM :)
At 4-bit quantization it should already fit quite nicely.