DeepSeek R1 NVFP4 — Hardware Requirements & GPU Compatibility
ChatReasoning
40.1K downloads
273 likesSpecifications
- Publisher
- NVIDIA
- Family
- DeepSeek R1
- Parameters
- 396.8B
- Architecture
- DeepseekV3ForCausalLM
- Context Length
- 163,840 tokens
- Vocabulary Size
- 129,280
- Release Date
- 2025-06-06
- License
- MIT
Get Started
HuggingFace
How Much VRAM Does DeepSeek R1 NVFP4 Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| IQ2_XXS | 2.20 | 113.0 GB | 396.0 GB | 109.11 GB | Importance-weighted 2-bit, extreme compression — significant quality loss |
| Q2_K | 3.40 | 172.5 GB | 455.5 GB | 168.63 GB | 2-bit quantization with K-quant improvements |
| Q3_K_M | 3.90 | 197.3 GB | 480.3 GB | 193.42 GB | 3-bit medium quantization |
| Q4_K_M | 4.80 | 241.9 GB | 524.9 GB | 238.06 GB | 4-bit medium quantization — most popular sweet spot |
| Q5_K_M | 5.70 | 286.6 GB | 569.5 GB | 282.70 GB | 5-bit medium quantization — good quality/size tradeoff |
| Q6_K | 6.60 | 331.2 GB | 614.2 GB | 327.33 GB | 6-bit quantization, very good quality |
| Q8_0 | 8.00 | 400.6 GB | 683.6 GB | 396.77 GB | 8-bit quantization, near-lossless |
Which GPUs Can Run DeepSeek R1 NVFP4?
Q4_K_M · 241.9 GBDeepSeek R1 NVFP4 (Q4_K_M) requires 241.9 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 315+ GB is recommended. Using the full 164K context window can add up to 283.0 GB, bringing total usage to 524.9 GB. No single GPU has enough memory — multi-GPU or cluster setups are needed.
Which Devices Can Run DeepSeek R1 NVFP4?
Q4_K_M · 241.9 GB2 devices with unified memory can run DeepSeek R1 NVFP4, including NVIDIA DGX H100, NVIDIA DGX A100 640GB.
Runs great
— Plenty of headroomRelated Models
Frequently Asked Questions
- How much VRAM does DeepSeek R1 NVFP4 need?
DeepSeek R1 NVFP4 requires 241.9 GB of VRAM at Q4_K_M, or 400.6 GB at Q8_0. Full 164K context adds up to 283.0 GB (524.9 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 396.8B × 4.8 bits ÷ 8 = 238.1 GB
KV Cache + Overhead ≈ 3.8 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 286.8 GB (at full 164K context)
VRAM usage by quantization
Q4_K_M241.9 GBQ4_K_M + full context524.9 GB- Can NVIDIA GeForce RTX 5090 run DeepSeek R1 NVFP4?
No — DeepSeek R1 NVFP4 requires at least 113.0 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
- What's the best quantization for DeepSeek R1 NVFP4?
For DeepSeek R1 NVFP4, Q4_K_M (241.9 GB) offers the best balance of quality and VRAM usage. Q5_K_M (286.6 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 113.0 GB.
VRAM requirement by quantization
IQ2_XXS113.0 GB~53%Q3_K_M197.3 GB~83%Q4_K_M ★241.9 GB~89%Q5_K_M286.6 GB~92%Q6_K331.2 GB~95%Q8_0400.6 GB~99%★ Recommended — best balance of quality and VRAM usage.
- Can I run DeepSeek R1 NVFP4 on a Mac?
DeepSeek R1 NVFP4 requires at least 113.0 GB at IQ2_XXS, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run DeepSeek R1 NVFP4 locally?
Yes — DeepSeek R1 NVFP4 can run locally on consumer hardware. At Q4_K_M quantization it needs 241.9 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- What's the download size of DeepSeek R1 NVFP4?
At Q4_K_M, the download is about 238.06 GB. The full-precision Q8_0 version is 396.77 GB. The smallest option (IQ2_XXS) is 109.11 GB.
- Which GPUs can run DeepSeek R1 NVFP4?
No single consumer GPU has enough VRAM to run DeepSeek R1 NVFP4 at Q4_K_M (241.9 GB). Multi-GPU or professional hardware is required.
- Which devices can run DeepSeek R1 NVFP4?
2 devices with unified memory can run DeepSeek R1 NVFP4 at Q4_K_M (241.9 GB), including NVIDIA DGX A100 640GB, NVIDIA DGX H100. Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.