Qwen3 30B A3B FP8 — Hardware Requirements & GPU Compatibility
Chat
Qwen3 30B A3B FP8 is the FP8 precision version of Alibaba's 30-billion-parameter mixture-of-experts model with approximately 3 billion active parameters per token. FP8 provides a good balance between quantization efficiency and output quality, sitting between full precision and more aggressive INT4 or INT8 formats. This variant is aimed at users who want near-original model quality with meaningful memory savings. The MoE architecture already keeps compute demands low, and FP8 further reduces the VRAM footprint, making it a practical choice for consumer GPUs in the 8 to 12 GB VRAM range.
87.6K downloads
82 likesSpecifications
- Publisher
- Alibaba
- Family
- Qwen
- Parameters
- 30B
- Architecture
- Qwen3MoeForCausalLM
- Context Length
- 40,960 tokens
- Vocabulary Size
- 151,936
- Release Date
- 2025-07-26
- License
- Apache 2.0
Get Started
HuggingFace
How Much VRAM Does Qwen3 30B A3B FP8 Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| IQ2_XXS | 2.20 | 8.7 GB | 10.6 GB | 8.25 GB | Importance-weighted 2-bit, extreme compression — significant quality loss |
| IQ2_XS | 2.40 | 9.4 GB | 11.3 GB | 9.00 GB | Importance-weighted 2-bit, extra small |
| IQ2_S | 2.50 | 9.8 GB | 11.7 GB | 9.38 GB | Importance-weighted 2-bit, small |
| IQ2_M | 2.70 | 10.5 GB | 12.4 GB | 10.13 GB | Importance-weighted 2-bit, medium |
| IQ3_XXS | 3.10 | 12.0 GB | 13.9 GB | 11.63 GB | Importance-weighted 3-bit |
| IQ3_XS | 3.30 | 12.8 GB | 14.7 GB | 12.38 GB | Importance-weighted 3-bit, extra small |
| Q2_K | 3.40 | 13.2 GB | 15.1 GB | 12.75 GB | 2-bit quantization with K-quant improvements |
| Q3_K_S | 3.50 | 13.5 GB | 15.4 GB | 13.13 GB | 3-bit small quantization |
| IQ3_M | 3.60 | 13.9 GB | 15.8 GB | 13.50 GB | Importance-weighted 3-bit, medium |
| Q3_K_M | 3.90 | 15.0 GB | 16.9 GB | 14.63 GB | 3-bit medium quantization |
| Q4_0 | 4.00 | 15.4 GB | 17.3 GB | 15.00 GB | 4-bit legacy quantization |
| Q3_K_L | 4.10 | 15.8 GB | 17.7 GB | 15.37 GB | 3-bit large quantization |
| IQ4_XS | 4.30 | 16.5 GB | 18.4 GB | 16.13 GB | Importance-weighted 4-bit, compact |
| Q4_1 | 4.50 | 17.3 GB | 19.2 GB | 16.88 GB | 4-bit legacy quantization with offset |
| Q4_K_S | 4.50 | 17.3 GB | 19.2 GB | 16.88 GB | 4-bit small quantization |
| IQ4_NL | 4.50 | 17.3 GB | 19.2 GB | 16.88 GB | Importance-weighted 4-bit, non-linear |
| Q4_K_M | 4.80 | 18.4 GB | 20.3 GB | 18.00 GB | 4-bit medium quantization — most popular sweet spot |
| Q4_K_L | 4.90 | 18.8 GB | 20.7 GB | 18.38 GB | 4-bit large quantization |
| Q5_K_S | 5.50 | 21.0 GB | 22.9 GB | 20.63 GB | 5-bit small quantization |
| Q5_K_M | 5.70 | 21.8 GB | 23.7 GB | 21.38 GB | 5-bit medium quantization — good quality/size tradeoff |
| Q5_K_L | 5.80 | 22.1 GB | 24.1 GB | 21.75 GB | 5-bit large quantization |
| Q6_K | 6.60 | 25.1 GB | 27.1 GB | 24.75 GB | 6-bit quantization, very good quality |
| Q8_0 | 8.00 | 30.4 GB | 32.3 GB | 30.00 GB | 8-bit quantization, near-lossless |
Which GPUs Can Run Qwen3 30B A3B FP8?
Q4_K_M · 18.4 GBQwen3 30B A3B FP8 (Q4_K_M) requires 18.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 24+ GB is recommended. Using the full 41K context window can add up to 1.9 GB, bringing total usage to 20.3 GB. 6 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti.
Runs great
— Plenty of headroom
Which Devices Can Run Qwen3 30B A3B FP8?
Q4_K_M · 18.4 GB21 devices with unified memory can run Qwen3 30B A3B FP8, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Mini M4 Pro (24 GB).
Runs great
— Plenty of headroomRelated Models
Frequently Asked Questions
- How much VRAM does Qwen3 30B A3B FP8 need?
Qwen3 30B A3B FP8 requires 18.4 GB of VRAM at Q4_K_M, or 30.4 GB at Q8_0. Full 41K context adds up to 1.9 GB (20.3 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 30B × 4.8 bits ÷ 8 = 18 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 2.3 GB (at full 41K context)
VRAM usage by quantization
Q4_K_M18.4 GBQ4_K_M + full context20.3 GB- Can NVIDIA GeForce RTX 4090 run Qwen3 30B A3B FP8?
Yes, at Q5_K_L (22.1 GB) or lower. Higher quantizations like Q6_K (25.1 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
- What's the best quantization for Qwen3 30B A3B FP8?
For Qwen3 30B A3B FP8, Q4_K_M (18.4 GB) offers the best balance of quality and VRAM usage. Q4_K_L (18.8 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 8.7 GB.
VRAM requirement by quantization
IQ2_XXS8.7 GB~53%Q2_K13.2 GB~75%Q3_K_L15.8 GB~86%Q4_K_M ★18.4 GB~89%Q4_K_L18.8 GB~90%Q8_030.4 GB~99%★ Recommended — best balance of quality and VRAM usage.
- Can I run Qwen3 30B A3B FP8 on a Mac?
Qwen3 30B A3B FP8 requires at least 8.7 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 FP8 locally?
Yes — Qwen3 30B A3B FP8 can run locally on consumer hardware. At Q4_K_M quantization it needs 18.4 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Qwen3 30B A3B FP8?
At Q4_K_M, Qwen3 30B A3B FP8 can reach ~158 tok/s on AMD Instinct MI300X. On NVIDIA GeForce RTX 4090: ~36 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 ÷ 18.4 × 0.55 = ~158 tok/s
Estimated speed at Q4_K_M (18.4 GB)
AMD Instinct MI300X~158 tok/sNVIDIA GeForce RTX 4090~36 tok/sNVIDIA H100 SXM~118 tok/sAMD Instinct MI250X~98 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 FP8?
At Q4_K_M, the download is about 18.00 GB. The full-precision Q8_0 version is 30.00 GB. The smallest option (IQ2_XXS) is 8.25 GB.