LLM EXPLORER 57,252 MODELS INDEXED

DeepSeek V3.2 Mlx 5bit by mlx-community

By mlx-community · 98 downloads

DeepSeek V3.2 Mlx 5bit is an open-source language model by mlx-community. Features: 671.9b LLM, VRAM: 241.6GB, Context: 160K, License: mit, Quantized, LLM Explorer Score: 0.22.

  5-bit   5bit Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-...   Deepseek v32   Mlx   Quantized   Region:us   Safetensors   Sharded   Tensorflow

DeepSeek V3.2 Mlx 5bit Parameters and Internals

LLM NameDeepSeek V3.2 Mlx 5bit
Repository 🤗https://huggingface.co/mlx-community/DeepSeek-V3.2-mlx-5bit 
Base Model(s)  deepseek-ai/DeepSeek-V3.2   deepseek-ai/DeepSeek-V3.2
Model Size671.9b
Required VRAM241.6 GB
Updated2026-08-01
Maintainermlx-community
Model Typedeepseek_v32
Model Files  4.6 GB: 1-of-88   5.3 GB: 2-of-88   5.2 GB: 3-of-88   5.3 GB: 4-of-88   5.3 GB: 5-of-88   5.2 GB: 6-of-88   5.3 GB: 7-of-88   5.3 GB: 8-of-88   5.2 GB: 9-of-88   5.3 GB: 10-of-88   5.3 GB: 11-of-88   5.2 GB: 12-of-88   5.3 GB: 13-of-88   5.3 GB: 14-of-88   5.2 GB: 15-of-88   5.3 GB: 16-of-88   5.3 GB: 17-of-88   5.2 GB: 18-of-88   5.3 GB: 19-of-88   5.3 GB: 20-of-88   5.2 GB: 21-of-88   5.3 GB: 22-of-88   5.3 GB: 23-of-88   5.2 GB: 24-of-88   5.3 GB: 25-of-88   5.3 GB: 26-of-88   5.2 GB: 27-of-88   5.3 GB: 28-of-88   5.3 GB: 29-of-88   5.2 GB: 30-of-88   5.3 GB: 31-of-88   5.3 GB: 32-of-88   5.2 GB: 33-of-88   5.3 GB: 34-of-88   5.3 GB: 35-of-88   5.2 GB: 36-of-88   5.3 GB: 37-of-88   5.3 GB: 38-of-88   5.2 GB: 39-of-88   5.3 GB: 40-of-88   5.3 GB: 41-of-88   5.2 GB: 42-of-88   5.3 GB: 43-of-88   5.3 GB: 44-of-88   5.2 GB: 45-of-88   5.3 GB: 46-of-88
Quantization Type5bit
Model ArchitectureDeepseekV32ForCausalLM
Licensemit
Context Length163840
Model Max Length163840
Transformers Version4.44.2
Tokenizer ClassLlamaTokenizer
Padding Token<|end▁of▁sentence|>
Vocabulary Size129280
Torch Data Typebfloat16

Best Alternatives to DeepSeek V3.2 Mlx 5bit

Best Alternatives
Context / RAM
Downloads
Likes
DeepSeek V3.2 4bit160K / 198.8 GB8467
Note: green Score (e.g. "73.2") means that the model is better than mlx-community/DeepSeek-V3.2-mlx-5bit.