Qwen3.5-27B
Files and memory at 8K context
| Quant | File | Files | + KV + runtime | Smallest fit | Source |
|---|---|---|---|---|---|
| Q4_K_M unsloth/Qwen3.5-27B-GGUF | 15.59 GiB | 1 | 17.1 GiB+ | 20 GB GPU | HF files |
| Q5_K_M unsloth/Qwen3.5-27B-GGUF | 18.26 GiB | 1 | 19.8 GiB+ | 24 GB GPU | HF files |
| Q8_0 unsloth/Qwen3.5-27B-GGUF | 26.63 GiB | 1 | 28.1 GiB+ | 32 GB GPU | HF files |
| MLX 4-bit mlx-community/Qwen3.5-27B-4bit | 14.95 GiB | 3 | 16.5 GiB+ | 24 GB Mac | HF files |
| MLX 8-bit mlx-community/Qwen3.5-27B-8bit | 27.48 GiB | 6 | 29.0 GiB+ | 48 GB Mac | HF files |
No verified file: FP16. Sizes are exact file bytes (text weights; projectors listed separately). “+” = recurrent state or other terms unknown. Smallest fit adds 0.5 GiB spare; Macs use 75% of memory.
Quality
Artificial Analysis Intelligence Index
Unknown
Arena text rating
Unknown
Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.
Reported speeds 10
| Hardware | Quant | Backend | Prompt t/s | Gen t/s | Date | Source |
|---|---|---|---|---|---|---|
| M5 Max (40c GPU) 128 GB | MLX 4bit | MLX | 943 | 32 | 2026-10-16 | omlx.ai |
| M1 Max (24c GPU) 32 GB | MLX 4bit | MLX | 77.3 | 14.8 | 2026-10-06 | omlx.ai |
| M4 Max (40c GPU) 48 GB | MLX 4bit | MLX | 262 | 27.9 | 2026-08-25 | omlx.ai |
| M2 Ultra (60c GPU) 64 GB | MLX 4bit | MLX | 191 | 30.8 | 2026-08-20 | omlx.ai |
| M5 Pro (20c GPU) 64 GB | MLX 4bit | MLX | 469 | 17.1 | 2026-06-29 | omlx.ai |
| M4 Pro (20c GPU) 48 GB | MLX 4bit | MLX | 117 | 14.5 | 2026-03-13 | omlx.ai |
| NVIDIA GeForce RTX 3090 24 GB | Q4_K (subvariant not stated) | llama.cpp CUDA | 1,104 | 33.5 | unknown | hardware-corner.net |
| NVIDIA GeForce RTX 3090 24 GB | Q4_K (subvariant not stated) | llama.cpp CUDA | 679 | 28.8 | unknown | hardware-corner.net |
| NVIDIA GeForce RTX 5090 32 GB | Q4_K (subvariant not stated) | llama.cpp CUDA | 3,004 | 58.8 | unknown | hardware-corner.net |
| NVIDIA GeForce RTX 5090 32 GB | Q4_K (subvariant not stated) | llama.cpp CUDA | 1,020 | 44.1 | unknown | hardware-corner.net |
Reported by the linked sources, not reproduced by LocalFit. All rows in the speed database
Community threads
- Chinese-English Typography Issues Hugging Face · 2026-03-13
- [transformers-server] Continuous batching does not support Qwen-3.5 series Hugging Face · 2026-03-09
Qualitative signals only; titles are not verified claims.