What are the hardware requirements for DeepSeek V4 Flash 2bit M DQ?

DeepSeek V4 Flash 2bit M DQ requires approximately 106.4 GB of VRAM and supports a context window of 1024K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed DeepSeek V4 Flash 2bit M DQ and how large is it?

DeepSeek V4 Flash 2bit M DQ is developed by mlx-community, a model with 284.3b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does DeepSeek V4 Flash 2bit M DQ perform on standard benchmarks?

DeepSeek V4 Flash 2bit M DQ has the following published scores: LMSYS ELO 1436. Compare against reference models on this page or on the LLM Explorer leaderboards.

DeepSeek V4 Flash 2bit M DQ by mlx-community — VRAM 106.4GB, 1024K context

Name: DeepSeek V4 Flash 2bit M DQ
Author: mlx-community

DeepSeek V4 Flash 2bit M DQ is an open-source language model by mlx-community. Features: 284.3b LLM, VRAM: 106.4GB, Context: 1024K, License: mit, Quantized, LLM Explorer Score: 0.41, ELO: 1436.

2bit 8-bit Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-... Conversational Deepseek v4 Mlx Quantized Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/mlx-community/DeepSeek-V4-Flash-2bit-M-DQ

DeepSeek V4 Flash 2bit M DQ 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").

What is the LLM Explorer Rank (Score)

DeepSeek V4 Flash 2bit M DQ Parameters and Internals

LLM Name	DeepSeek V4 Flash 2bit M DQ
Repository 🤗	https://huggingface.co/mlx-community/DeepSeek-V4-Flash-2bit-M-DQ
Base Model(s)	DeepSeek V4 Flash deepseek-ai/DeepSeek-V4-Flash
Model Size	284.3b
Required VRAM	106.4 GB
Updated	2026-08-06
Maintainer	mlx-community
Model Type	deepseek_v4
Model Files	5.3 GB: 1-of-22 5.1 GB: 2-of-22 4.9 GB: 3-of-22 4.9 GB: 4-of-22 4.9 GB: 5-of-22 4.9 GB: 6-of-22 4.9 GB: 7-of-22 4.9 GB: 8-of-22 4.9 GB: 9-of-22 4.9 GB: 10-of-22 4.9 GB: 11-of-22 4.9 GB: 12-of-22 4.9 GB: 13-of-22 4.9 GB: 14-of-22 4.9 GB: 15-of-22 4.9 GB: 16-of-22 4.9 GB: 17-of-22 4.9 GB: 18-of-22 4.9 GB: 19-of-22 4.9 GB: 20-of-22 4.9 GB: 21-of-22 2.9 GB: 22-of-22
Quantization Type	2bit
Model Architecture	DeepseekV4ForCausalLM
License	mit
Context Length	1048576
Model Max Length	1048576
Transformers Version	4.57.1
Tokenizer Class	TokenizersBackend
Padding Token	<｜end▁of▁sentence｜>
Vocabulary Size	129280
Torch Data Type	bfloat16

Rank the DeepSeek V4 Flash 2bit M DQ capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to DeepSeek V4 Flash 2bit M DQ

Best Alternatives	Context / RAM	Downloads	Likes
DeepSeek V4 Flash 4bit	1024K / 152.2 GB	12635	25
DeepSeek V4 Flash 3bit DQ	1024K / 128.9 GB	696	2
DeepSeek V4 Flash 2bit DQ	1024K / 96.3 GB	14425	47
...seek Ai DeepSeek V4 Flash 8bit	1024K / 215.2 GB	3577	12
...seek Ai DeepSeek V4 Flash 4bit	1024K / 159.7 GB	1223	5
DeepSeek V4 Flash 6bit	1024K / 153.5 GB	781	2
DeepSeek V4 Flash BF16	1024K / 568.7 GB	360	8
DeepSeek V4 Flash Mxfp8	1024K / 154.8 GB	808	2

Note: green Score (e.g. "73.2") means that the model is better than mlx-community/DeepSeek-V4-Flash-2bit-M-DQ.