LLM EXPLORER 58,531 MODELS INDEXED

Deepseek Ai DeepSeek V4 Flash 6bit by mlx-community

By mlx-community · 367 downloads
HF Repo ↗ Copy URL

Deepseek Ai DeepSeek V4 Flash 6bit is an open-source language model by mlx-community. Features: 284.3b LLM, VRAM: 230.9GB, Context: 1024K, License: mit, Quantized, LLM Explorer Score: 0.4, ELO: 1436.

  6-bit   6bit Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-...   Deepseek v4   Mlx   Quantized   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/mlx-community/deepseek-ai-DeepSeek-V4-Flash-6bit 

Deepseek Ai DeepSeek V4 Flash 6bit Benchmarks

LMSys ELO 1436 vs 1272 (so35) 12.9%
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

Deepseek Ai DeepSeek V4 Flash 6bit Parameters and Internals

LLM NameDeepseek Ai DeepSeek V4 Flash 6bit
Repository 🤗https://huggingface.co/mlx-community/deepseek-ai-DeepSeek-V4-Flash-6bit 
Base Model(s)  DeepSeek V4 Flash   deepseek-ai/DeepSeek-V4-Flash
Model Size284.3b
Required VRAM230.9 GB
Updated2026-08-10
Maintainermlx-community
Model Typedeepseek_v4
Model Files  4.0 GB: 1-of-44   5.3 GB: 2-of-44   5.3 GB: 3-of-44   5.3 GB: 4-of-44   5.4 GB: 5-of-44   5.3 GB: 6-of-44   5.4 GB: 7-of-44   5.3 GB: 8-of-44   5.4 GB: 9-of-44   5.3 GB: 10-of-44   5.4 GB: 11-of-44   5.3 GB: 12-of-44   5.4 GB: 13-of-44   5.3 GB: 14-of-44   5.4 GB: 15-of-44   5.3 GB: 16-of-44   5.4 GB: 17-of-44   5.3 GB: 18-of-44   5.4 GB: 19-of-44   5.3 GB: 20-of-44   5.4 GB: 21-of-44   5.3 GB: 22-of-44   5.4 GB: 23-of-44   5.3 GB: 24-of-44   5.4 GB: 25-of-44   5.3 GB: 26-of-44   5.4 GB: 27-of-44   5.3 GB: 28-of-44   5.4 GB: 29-of-44   5.3 GB: 30-of-44   5.4 GB: 31-of-44   5.3 GB: 32-of-44   5.4 GB: 33-of-44   5.3 GB: 34-of-44   5.4 GB: 35-of-44   5.3 GB: 36-of-44   5.4 GB: 37-of-44   5.3 GB: 38-of-44   5.4 GB: 39-of-44   5.3 GB: 40-of-44   5.4 GB: 41-of-44   5.3 GB: 42-of-44   5.4 GB: 43-of-44   2.3 GB: 44-of-44
Quantization Type6bit
Model ArchitectureDeepseekV4ForCausalLM
Licensemit
Context Length1048576
Model Max Length1048576
Transformers Version4.57.1
Vocabulary Size129280
Torch Data Typebfloat16

Best Alternatives to Deepseek Ai DeepSeek V4 Flash 6bit

Best Alternatives
Context / RAM
Downloads
Likes
DeepSeek V4 Flash 4bit1024K / 152.2 GB1263525
DeepSeek V4 Flash 2bit M DQ1024K / 106.4 GB7852
DeepSeek V4 Flash 3bit DQ1024K / 128.9 GB6962
...seek Ai DeepSeek V4 Flash 3bit1024K / 124.3 GB4771
DeepSeek V4 Flash 2bit DQ1024K / 96.3 GB1442547
...seek Ai DeepSeek V4 Flash 8bit1024K / 215.2 GB357712
...seek Ai DeepSeek V4 Flash 4bit1024K / 159.7 GB12235
DeepSeek V4 Flash 6bit1024K / 153.5 GB7812
DeepSeek V4 Flash 8bit1024K / 154.8 GB5721
...eek V4 Flash 0731 GPTQ Int4 V11024K / 153 GB252
Note: green Score (e.g. "73.2") means that the model is better than mlx-community/deepseek-ai-DeepSeek-V4-Flash-6bit.