Qwen3 30B A3B NVFP4 — Hardware Requirements & GPU Compatibility
Chat
Qwen3 30B A3B NVFP4 is NVIDIA's NVFP4-quantized version of Alibaba's Qwen3 30B mixture-of-experts model, compressed to roughly 15.6 billion parameters of effective memory usage. With only 3 billion parameters active per token, it runs remarkably fast for a model of its intelligence class. This is one of the most efficient models available for local deployment. The combination of MoE architecture and NVFP4 quantization means you get 30B-class reasoning and instruction-following on hardware that would normally struggle with models half this size. It's an excellent choice for users who want strong performance without top-tier GPUs.
79.4K downloads
25 likesSpecifications
- Publisher
- NVIDIA
- Family
- Qwen
- Parameters
- 15.6B
- Architecture
- Qwen3MoeForCausalLM
- Context Length
- 40,960 tokens
- Vocabulary Size
- 151,936
- Release Date
- 2025-09-10
- License
- Apache 2.0
Get Started
HuggingFace
How Much VRAM Does Qwen3 30B A3B NVFP4 Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| IQ2_XXS | 2.20 | 4.7 GB | 6.6 GB | 4.29 GB | Importance-weighted 2-bit, extreme compression — significant quality loss |
| IQ2_XS | 2.40 | 5.1 GB | 7.0 GB | 4.68 GB | Importance-weighted 2-bit, extra small |
| IQ2_S | 2.50 | 5.3 GB | 7.2 GB | 4.87 GB | Importance-weighted 2-bit, small |
| IQ2_M | 2.70 | 5.7 GB | 7.6 GB | 5.26 GB | Importance-weighted 2-bit, medium |
| IQ3_XXS | 3.10 | 6.4 GB | 8.3 GB | 6.04 GB | Importance-weighted 3-bit |
| IQ3_XS | 3.30 | 6.8 GB | 8.7 GB | 6.43 GB | Importance-weighted 3-bit, extra small |
| Q2_K | 3.40 | 7.0 GB | 8.9 GB | 6.62 GB | 2-bit quantization with K-quant improvements |
| Q3_K_S | 3.50 | 7.2 GB | 9.1 GB | 6.82 GB | 3-bit small quantization |
| IQ3_M | 3.60 | 7.4 GB | 9.3 GB | 7.01 GB | Importance-weighted 3-bit, medium |
| Q3_K_M | 3.90 | 8 GB | 9.9 GB | 7.60 GB | 3-bit medium quantization |
| Q4_0 | 4.00 | 8.2 GB | 10.1 GB | 7.79 GB | 4-bit legacy quantization |
| Q3_K_L | 4.10 | 8.4 GB | 10.3 GB | 7.99 GB | 3-bit large quantization |
| IQ4_XS | 4.30 | 8.8 GB | 10.7 GB | 8.38 GB | Importance-weighted 4-bit, compact |
| Q4_1 | 4.50 | 9.2 GB | 11.1 GB | 8.77 GB | 4-bit legacy quantization with offset |
| Q4_K_S | 4.50 | 9.2 GB | 11.1 GB | 8.77 GB | 4-bit small quantization |
| IQ4_NL | 4.50 | 9.2 GB | 11.1 GB | 8.77 GB | Importance-weighted 4-bit, non-linear |
| Q4_K_M | 4.80 | 9.8 GB | 11.7 GB | 9.35 GB | 4-bit medium quantization — most popular sweet spot |
| Q4_K_L | 4.90 | 9.9 GB | 11.9 GB | 9.54 GB | 4-bit large quantization |
| Q5_K_S | 5.50 | 11.1 GB | 13.0 GB | 10.71 GB | 5-bit small quantization |
| Q5_K_M | 5.70 | 11.5 GB | 13.4 GB | 11.10 GB | 5-bit medium quantization — good quality/size tradeoff |
| Q5_K_L | 5.80 | 11.7 GB | 13.6 GB | 11.30 GB | 5-bit large quantization |
| Q6_K | 6.60 | 13.3 GB | 15.2 GB | 12.86 GB | 6-bit quantization, very good quality |
| Q8_0 | 8.00 | 16.0 GB | 17.9 GB | 15.58 GB | 8-bit quantization, near-lossless |
Which GPUs Can Run Qwen3 30B A3B NVFP4?
Q4_K_M · 9.8 GBQwen3 30B A3B NVFP4 (Q4_K_M) requires 9.8 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 13+ GB is recommended. Using the full 41K context window can add up to 1.9 GB, bringing total usage to 11.7 GB. 28 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3080 Ti.
Runs great
— Plenty of headroom
Decent
— Enough VRAM, may be tightWhich Devices Can Run Qwen3 30B A3B NVFP4?
Q4_K_M · 9.8 GB27 devices with unified memory can run Qwen3 30B A3B NVFP4, including NVIDIA DGX H100, NVIDIA DGX A100 640GB.
Runs great
— Plenty of headroomRelated Models
Frequently Asked Questions
- How much VRAM does Qwen3 30B A3B NVFP4 need?
Qwen3 30B A3B NVFP4 requires 9.8 GB of VRAM at Q4_K_M, or 16.0 GB at Q8_0. Full 41K context adds up to 1.9 GB (11.7 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 15.6B × 4.8 bits ÷ 8 = 9.4 GB
KV Cache + Overhead ≈ 0.3 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 2.3 GB (at full 41K context)
VRAM usage by quantization
Q4_K_M9.8 GBQ4_K_M + full context11.7 GB- What's the best quantization for Qwen3 30B A3B NVFP4?
For Qwen3 30B A3B NVFP4, Q4_K_M (9.8 GB) offers the best balance of quality and VRAM usage. Q4_K_L (9.9 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 4.7 GB.
VRAM requirement by quantization
IQ2_XXS4.7 GB~53%Q2_K7.0 GB~75%Q3_K_L8.4 GB~86%Q4_K_M ★9.8 GB~89%Q4_K_L9.9 GB~90%Q8_016.0 GB~99%★ Recommended — best balance of quality and VRAM usage.
- Can I run Qwen3 30B A3B NVFP4 on a Mac?
Qwen3 30B A3B NVFP4 requires at least 4.7 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 Qwen3 30B A3B NVFP4 locally?
Yes — Qwen3 30B A3B NVFP4 can run locally on consumer hardware. At Q4_K_M quantization it needs 9.8 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Qwen3 30B A3B NVFP4?
At Q4_K_M, Qwen3 30B A3B NVFP4 can reach ~299 tok/s on AMD Instinct MI300X. On NVIDIA GeForce RTX 4090: ~67 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: AMD Instinct MI300X → 5300 ÷ 9.8 × 0.55 = ~299 tok/s
Estimated speed at Q4_K_M (9.8 GB)
AMD Instinct MI300X~299 tok/sNVIDIA GeForce RTX 4090~67 tok/sNVIDIA H100 SXM~224 tok/sAMD Instinct MI250X~185 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Qwen3 30B A3B NVFP4?
At Q4_K_M, the download is about 9.35 GB. The full-precision Q8_0 version is 15.58 GB. The smallest option (IQ2_XXS) is 4.29 GB.