Qwen-3.5-unsloth-mlx
Collection
AWQ-style pre-scaling using Unsloth's imatrix calibration data, then 3-6-bit affine quantization with the Unsloth mixed-precision recipe via MLX • 20 items • Updated • 20
Qwen3.5-27B — UD-Q6_K_XL (mlx-node)
6-bit base mixed-precision quantization of Qwen/Qwen3.5-27B for Apple Silicon via mlx-node.
| Original (BF16) | This Model | |
|---|---|---|
| Size | ~51 GB | 27 GB |
| Precision | BF16 uniform | Mixed 6/8/8/8/8-bit + BF16 |
All Variants
| Repo | GGUF Equivalent | Size |
|---|---|---|
| Brooooooklyn/Qwen3.5-27B-UD-Q2_K_XL-mlx | UD-Q2_K_XL | 15 GB |
| Brooooooklyn/Qwen3.5-27B-UD-Q3_K_XL-mlx | UD-Q3_K_XL | 17 GB |
| Brooooooklyn/Qwen3.5-27B-UD-Q4_K_XL-mlx | UD-Q4_K_XL | 20 GB |
| Brooooooklyn/Qwen3.5-27B-UD-Q5_K_XL-mlx | UD-Q5_K_XL | 24 GB |
| Brooooooklyn/Qwen3.5-27B-UD-Q6_K_XL-mlx | UD-Q6_K_XL | 27 GB |
| Brooooooklyn/Qwen3.5-27B-UD-Q8_K_XL-mlx | UD-Q8_K_XL | 29 GB |
Per-Tensor Bit Assignments (N=6)
| Weight | Bits |
|---|---|
| embed_tokens | 8-bit |
| lm_head | 8-bit |
| self_attn.q/k/v_proj | 8-bit + AWQ |
| linear_attn.in_proj_qkv/z | 8-bit + AWQ |
| self_attn.o_proj | bf16 |
| linear_attn.out_proj | bf16 |
| down_proj | 8-bit |
| gate_proj, up_proj | 6-bit |
| Based on Unsloth Dynamic 2.0. Apache 2.0. |
- Downloads last month
- 368
Model size
9B params
Tensor type
BF16
·
U32 ·
Hardware compatibility
Log In to add your hardware
6-bit
Model tree for Brooooooklyn/Qwen3.5-27B-UD-Q6_K_XL-mlx
Base model
Qwen/Qwen3.5-27B