Nanbeige4.2-3B Date unverified
Files and memory at 8K context
| Quant | File | Files | + KV + runtime | Smallest fit | Source |
|---|---|---|---|---|---|
| Q4_K_M mradermacher/Nanbeige4.2-3B-GGUF | 2.40 GiB | 1 | unknown | 8 GB GPU (cache unknown) | HF files |
| Q5_K_M mradermacher/Nanbeige4.2-3B-GGUF | 2.78 GiB | 1 | unknown | 8 GB GPU (cache unknown) | HF files |
| Q8_0 mradermacher/Nanbeige4.2-3B-GGUF | 4.13 GiB | 1 | unknown | 8 GB GPU (cache unknown) | HF files |
| FP16 mradermacher/Nanbeige4.2-3B-GGUF | 7.77 GiB | 1 | unknown | 12 GB GPU (cache unknown) | HF files |
| MLX 4-bit MercuriusDream/Nanbeige4.2-3B-mlx-4bit | 2.18 GiB | 1 | unknown | 8 GB Mac (cache unknown) | HF files |
| MLX 8-bit MercuriusDream/Nanbeige4.2-3B-mlx-8bit | 4.13 GiB | 1 | unknown | 8 GB Mac (cache unknown) | HF files |
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 2
| Hardware | Quant | Backend | Prompt t/s | Gen t/s | Date | Source |
|---|---|---|---|---|---|---|
| Steam Deck (AMD Custom GPU 0932 / RADV VANGOGH, Vulkan) | Q4_K_M | llama.cpp Vulkan | 124 | 15.3 | 2026-08-20 | llama.cpp #10879 |
| Steam Deck (AMD Custom GPU 0932 / RADV VANGOGH, Vulkan) | Q4_K_M | llama.cpp Vulkan | 126 | 16.4 | 2026-08-20 | llama.cpp #10879 |
Reported by the linked sources, not reproduced by LocalFit. All rows in the speed database
Community threads
- Q4_K_M slower than Q8_0 Hugging Face · 2026-09-10
- Why is it much slower and much more memory hungry than slightly larger models such as Qwen3.5-4B? Hugging Face · 2026-08-13
Qualitative signals only; titles are not verified claims.