DeepSeek V4 Flash W4A16 FP8 is an open-source language model by pastapaul. Features: 44.1b LLM, VRAM: 152.5GB, Context: 1024K, License: apache-2.0, LLM Explorer Score: 0.36.
Base model:deepseek-ai/deepsee... Base model:quantized:deepseek-... Compressed-tensors Deepseek Deepseek-v4 Deepseek v4 En Fp8-block Gptq Mixture-of-experts Moe Region:us Safetensors Sharded Tensorflow Vllm W4a16 Zh
Model Card on HF 🤗: https://huggingface.co/pastapaul/DeepSeek-V4-Flash-W4A16-FP8
DeepSeek V4 Flash W4A16 FP8 Benchmarks
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 W4A16 FP8 Parameters and Internals
| LLM Name | DeepSeek V4 Flash W4A16 FP8 |
| Repository 🤗 | https://huggingface.co/pastapaul/DeepSeek-V4-Flash-W4A16-FP8 |
| Base Model(s) | |
| Model Size | 44.1b |
| Required VRAM | 152.5 GB |
| Updated | 2026-05-10 |
| Maintainer | pastapaul |
| Model Type | deepseek_v4 |
| Model Files | |
| Supported Languages | en zh |
| Model Architecture | DeepseekV4ForCausalLM |
| License | apache-2.0 |
| Context Length | 1048576 |
| Model Max Length | 1048576 |
| Transformers Version | 5.8.0.dev0 |
| Tokenizer Class | TokenizersBackend |
| Padding Token | <｜end▁of▁sentence｜> |
| Vocabulary Size | 129280 |
| Torch Data Type | bfloat16 |
Rank the DeepSeek V4 Flash W4A16 FP8 Capabilities
🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟
What open-source LLMs or SLMs are you in search of? 53620 in total.
Original data from HuggingFace, OpenCompass and various public git repos.
Check out Ag3ntum — our secure, self-hosted AI agent for server management.
Release v20260328a