LLM EXPLORER 65,444 MODELS INDEXED

DeepSeek V2.5 1210 4bit by mlx-community

By mlx-community · 305 downloads
HF Repo ↗ Copy URL

DeepSeek V2.5 1210 4bit is an open-source language model by mlx-community. Features: 36.9b LLM, VRAM: 133.1GB, Context: 160K, License: other, Quantized, LLM Explorer Score: 0.16.

  4-bit   4bit Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-...   Conversational   Custom code   Deepseek v2   Endpoints compatible   Mlx   Quantized   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/mlx-community/DeepSeek-V2.5-1210-4bit 

DeepSeek V2.5 1210 4bit Parameters and Internals

LLM NameDeepSeek V2.5 1210 4bit
Repository 🤗https://huggingface.co/mlx-community/DeepSeek-V2.5-1210-4bit 
Base Model(s)  deepseek-ai/DeepSeek-V2.5-1210   deepseek-ai/DeepSeek-V2.5-1210
Model Size36.9b
Required VRAM133.1 GB
Updated2026-05-21
Maintainermlx-community
Model Typedeepseek_v2
Model Files  5.0 GB: 1-of-26   5.2 GB: 2-of-26   5.2 GB: 3-of-26   5.3 GB: 4-of-26   5.2 GB: 5-of-26   5.2 GB: 6-of-26   5.3 GB: 7-of-26   5.2 GB: 8-of-26   5.2 GB: 9-of-26   5.3 GB: 10-of-26   5.2 GB: 11-of-26   5.2 GB: 12-of-26   5.3 GB: 13-of-26   5.2 GB: 14-of-26   5.2 GB: 15-of-26   5.3 GB: 16-of-26   5.2 GB: 17-of-26   5.2 GB: 18-of-26   5.3 GB: 19-of-26   5.2 GB: 20-of-26   5.2 GB: 21-of-26   5.3 GB: 22-of-26   5.2 GB: 23-of-26   5.2 GB: 24-of-26   5.3 GB: 25-of-26   2.5 GB: 26-of-26
Quantization Type4bit
Model ArchitectureDeepseekV2ForCausalLM
Licenseother
Context Length163840
Model Max Length163840
Transformers Version4.39.3
Tokenizer ClassLlamaTokenizer
Padding Token<｜end▁of▁sentence｜>
Vocabulary Size102400
Torch Data Typebfloat16