Nanbeige4.1-3B
Files and memory at 8K context
| Quant | File | Files | + KV + runtime | Smallest fit | Source |
|---|---|---|---|---|---|
| Q4_K_M mradermacher/Nanbeige4.1-3B-GGUF | 2.28 GiB | 1 | 3.8 GiB | 8 GB GPU | HF files |
| Q5_K_M mradermacher/Nanbeige4.1-3B-GGUF | 2.63 GiB | 1 | 4.1 GiB | 8 GB GPU | HF files |
| Q8_0 mradermacher/Nanbeige4.1-3B-GGUF | 3.90 GiB | 1 | 5.4 GiB | 8 GB GPU | HF files |
| FP16 mradermacher/Nanbeige4.1-3B-GGUF | 7.33 GiB | 1 | 8.8 GiB | 12 GB GPU | HF files |
| MLX 4-bit mlx-community/Nanbeige4.1-3B-4bit | 2.06 GiB | 1 | 3.6 GiB | 8 GB Mac | HF files |
| MLX 8-bit mlx-community/Nanbeige4.1-3B-8bit | 3.89 GiB | 1 | 5.4 GiB | 8 GB Mac | 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
8.4 AA estimate
Variant: Nanbeige4.1-3B · observed 2026-10-17 · source
Rank #69 of 75 researched 2026 models.
Arena text rating
Unknown
Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.
Reported speeds 0
No reported measurement yet. Submit one.
Community threads
- vLLM openai docker - tool config Hugging Face · 2026-02-18
- Training Data and inference scripts with tool calling , websearch and so on plus training scripts Hugging Face · 2026-02-16
Qualitative signals only; titles are not verified claims.