What are the hardware requirements for DeepSeek V4 Flash Vision Exp AWQ DSpark?

DeepSeek V4 Flash Vision Exp AWQ DSpark requires approximately 136.2 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 Vision Exp AWQ DSpark and how large is it?

DeepSeek V4 Flash Vision Exp AWQ DSpark is developed by Solstice-AI, a model with 55b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate DeepSeek V4 Flash Vision Exp AWQ DSpark?

DeepSeek V4 Flash Vision Exp AWQ DSpark is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

DeepSeek V4 Flash Vision Exp AWQ DSpark by Solstice-AI — VRAM 136.2GB, 1024K context

Name: DeepSeek V4 Flash Vision Exp AWQ DSpark
Author: Solstice-AI

DeepSeek V4 Flash Vision Exp AWQ DSpark is an open-source language model by Solstice-AI. Features: 55b LLM, VRAM: 136.2GB, Context: 1024K, License: mit, Quantized, LLM Explorer Score: 0.33.

4-bit Auto-round Awq Base model:quantized:unsloth/d... Base model:unsloth/deepseek-v4... Conversational Deepseek Deepseek-v4 Deepseek v4 Dspark En Gguf Image-text-to-text Int4 Long-context Multimodal Quantized Region:us Safetensors Sglang Sharded Solstice-ai Speculative-decoding Tensorflow Unsloth Vision Vllm W4a16 Zh

Model Card on HF 🤗: https://huggingface.co/Solstice-AI/DeepSeek-V4-Flash-Vision-Exp-AWQ-DSpark

DeepSeek V4 Flash Vision Exp AWQ DSpark Parameters and Internals

LLM Name	DeepSeek V4 Flash Vision Exp AWQ DSpark
Repository 🤗	https://huggingface.co/Solstice-AI/DeepSeek-V4-Flash-Vision-Exp-AWQ-DSpark
Base Model(s)	DeepSeek V4 Flash Vision Exp unsloth/DeepSeek-V4-Flash-Vision-Exp
Model Size	55b
Required VRAM	136.2 GB
Updated	2026-09-10
Maintainer	Solstice-AI
Model Type	deepseek_v4
Model Files	0.9 GB 2.0 GB: 1-of-48 3.5 GB: 2-of-48 3.5 GB: 3-of-48 3.5 GB: 4-of-48 3.5 GB: 5-of-48 3.5 GB: 6-of-48 3.5 GB: 7-of-48 3.5 GB: 8-of-48 3.5 GB: 9-of-48 3.5 GB: 10-of-48 3.5 GB: 11-of-48 3.5 GB: 12-of-48 3.5 GB: 13-of-48 3.5 GB: 14-of-48 3.5 GB: 15-of-48 3.5 GB: 16-of-48 3.5 GB: 17-of-48 3.5 GB: 18-of-48 3.5 GB: 19-of-48 3.5 GB: 20-of-48 3.5 GB: 21-of-48 3.5 GB: 22-of-48 3.5 GB: 23-of-48 3.5 GB: 24-of-48 3.5 GB: 25-of-48 3.5 GB: 26-of-48 3.5 GB: 27-of-48 3.5 GB: 28-of-48 3.5 GB: 29-of-48 1.1 GB: 30-of-48 3.5 GB: 31-of-48 3.5 GB: 32-of-48 3.5 GB: 33-of-48 3.5 GB: 34-of-48 3.5 GB: 35-of-48 3.6 GB: 36-of-48 3.5 GB: 37-of-48 3.5 GB: 38-of-48 3.5 GB: 39-of-48 3.5 GB: 40-of-48
Supported Languages	en zh
GGUF Quantization	Yes
AWQ Quantization	Yes
Quantization Type	awq\|gguf
Model Architecture	DeepseekV4ForCausalLM
License	mit
Context Length	1048576
Model Max Length	1048576
Transformers Version	5.16.0.dev0
Vocabulary Size	129280
Torch Data Type	bfloat16

Rank the DeepSeek V4 Flash Vision Exp AWQ DSpark 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