LLM EXPLORER 60,828 MODELS INDEXED

DeepSeek V4 Flash JANGTQ2 by OsaurusAI

By OsaurusAI · 2032 downloads
HF Repo ↗ Copy URL

DeepSeek V4 Flash JANGTQ2 is an open-source language model by OsaurusAI. Features: 20.2b LLM, VRAM: 45.1GB, Context: 1024K, License: mit, LLM Explorer Score: 0.28.

  Apple-silicon Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-...   Deepseek   Deepseek-v4   Deepseek-v4-flash   Deepseek v4   Hash-layers   Jang   Jangtq   Jangtq-prestack   Jangtq2   Mla   Mlx   Moe   Mtp   Mxtq   Osaurus   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/OsaurusAI/DeepSeek-V4-Flash-JANGTQ2 

DeepSeek V4 Flash JANGTQ2 Parameters and Internals

LLM NameDeepSeek V4 Flash JANGTQ2
Repository 🤗https://huggingface.co/OsaurusAI/DeepSeek-V4-Flash-JANGTQ2 
Base Model(s)  DeepSeek V4 Flash   deepseek-ai/DeepSeek-V4-Flash
Model Size20.2b
Required VRAM45.1 GB
Updated2026-08-07
MaintainerOsaurusAI
Model Typedeepseek_v4
Model Files  0.0 GB   1.1 GB: 1-of-75   1.1 GB: 2-of-75   1.1 GB: 3-of-75   1.1 GB: 4-of-75   1.1 GB: 5-of-75   1.1 GB: 6-of-75   1.1 GB: 7-of-75   1.1 GB: 8-of-75   1.1 GB: 9-of-75   1.1 GB: 10-of-75   1.1 GB: 11-of-75   1.1 GB: 12-of-75   1.1 GB: 13-of-75   1.1 GB: 14-of-75   1.1 GB: 15-of-75   1.1 GB: 16-of-75   1.1 GB: 17-of-75   1.1 GB: 18-of-75   1.1 GB: 19-of-75   1.1 GB: 20-of-75   1.1 GB: 21-of-75   1.1 GB: 22-of-75   1.1 GB: 23-of-75   1.1 GB: 24-of-75   1.1 GB: 25-of-75   1.1 GB: 26-of-75   1.1 GB: 27-of-75   1.1 GB: 28-of-75   1.1 GB: 29-of-75   1.1 GB: 30-of-75   1.1 GB: 31-of-75   1.1 GB: 32-of-75   1.1 GB: 33-of-75   1.1 GB: 34-of-75   1.1 GB: 35-of-75   1.1 GB: 36-of-75   1.1 GB: 37-of-75   1.1 GB: 38-of-75   1.1 GB: 39-of-75   1.1 GB: 40-of-75   1.1 GB: 41-of-75
Model ArchitectureDeepseekV4ForCausalLM
Licensemit
Context Length1048576
Model Max Length1048576
Transformers Version4.57.1
Vocabulary Size129280
Torch Data Typebfloat16