Qwen · Over 80B · Mixture of experts
Qwen3 Next 80B A3B Instruct
81.32 billion parameters in total, but only 3.19 billion are read for each token. That split is the whole point of the design: it costs the memory of a large model and the speed of a small one. Most of its layers are linear-attention layers that keep a small fixed state, so only 12 of 48 grow with the window and long contexts cost far less than on a conventional model.
About Qwen3 Next 80B A3B Instruct
Qwen3 Next 80B A3B Instruct is a mixture-of-experts model with 81.32B parameters, 3.19B of them active for each token, built from 48 layers with a hidden size of 2048. Its repository was published on 2025-09-09. The license listed on Hugging Face is apache-2.0. Most of its layers are linear-attention layers that keep a small fixed state, so only 12 of 48 grow with the window and long contexts cost far less than on a conventional model.
Memory needed, by quantization
At an 8k context window. Quality is our rough ranking of how much the compression costs you: anything at or above Q5 is hard to tell apart from the original in normal use.
| Quantization | File size | KV cache | Total VRAM | Quality | What it costs you |
|---|---|---|---|---|---|
| Q8_0 | 80.47 GB | 0.26 GB | 80.76 GB | 99% | Lossless in practice. Use it when the memory is there. |
| Q6_K | 62.1 GB | 0.26 GB | 62.46 GB | 98% | Very close to Q8 for two thirds of the size. |
| Q5_K_M | 53.87 GB | 0.26 GB | 54.26 GB | 96% | The quality-first choice when Q6 will not fit. |
| Q4_K_M | 45.73 GB | 0.26 GB | 46.15 GB | 93% | The default. Best size-to-quality ratio for local use. |
| IQ4_XS | 40.23 GB | 0.26 GB | 40.68 GB | 91% | Importance-matrix 4-bit. Q4_K_S quality, smaller file. |
| Q3_K_M | 37.02 GB | 0.26 GB | 37.47 GB | 86% | Degradation starts to show. A way to fit one size up. |
| IQ3_XXS | 28.97 GB | 0.26 GB | 29.46 GB | 79% | Aggressive. Only worth it on very large models. |
What a longer conversation costs
Same model at Q4_K_M, only the context window changes. This model mixes 36 linear-attention layers, which keep only a small fixed state, with 12 ordinary layers, so its cache grows far more slowly than a conventional model of the same depth.
| Context | KV cache | Total VRAM | Fits in 8 GB | Fits in 12 GB | Fits in 24 GB |
|---|---|---|---|---|---|
| 4k | 0.17 GB | 46.05 GB | no | no | no |
| 8k | 0.26 GB | 46.15 GB | no | no | no |
| 16k | 0.45 GB | 46.34 GB | no | no | no |
| 32k | 0.82 GB | 46.71 GB | no | no | no |
| 64k | 1.57 GB | 47.46 GB | no | no | no |
| 128k | 3.07 GB | 48.96 GB | no | no | no |
The "fits" columns allow for the roughly 0.8 GB Windows keeps for the desktop.
Which hardware runs Qwen3 Next 80B A3B Instruct
13 of 118 consumer devices run it at a quantization worth using, at an 8k context window. Another 10 can load it only by compressing the weights far enough to damage the model, marked with a warning below.
| Device | Memory | Verdict | Quantization | Used | Speed |
|---|---|---|---|---|---|
| RTX PRO 6000 Blackwell NVIDIA | 96 GB | Runs great | Q8_0 | 80.76 GB | 224.6 t/s |
| Apple M3 Ultra 256GB Apple | 256 GB | Runs great | Q8_0 | 81.04 GB | 94.9 t/s |
| Apple M3 Ultra 512GB Apple | 512 GB | Runs great | Q8_0 | 81.04 GB | 94.9 t/s |
| Apple M1 Ultra 128GB Apple | 128 GB | Runs great | Q8_0 | 81.04 GB | 92.7 t/s |
| Apple M2 Ultra 128GB Apple | 128 GB | Runs great | Q8_0 | 81.04 GB | 92.7 t/s |
| Apple M2 Ultra 192GB Apple | 192 GB | Runs great | Q8_0 | 81.04 GB | 92.7 t/s |
| Apple M4 Max 128GB Apple | 128 GB | Runs great | Q8_0 | 81.04 GB | 63.2 t/s |
| Apple M3 Max 128GB Apple | 128 GB | Runs great | Q8_0 | 81.04 GB | 46.3 t/s |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Runs great | Q8_0 | 81.07 GB | 34.2 t/s |
| Ryzen AI Max+ 395 128GB AMD | 128 GB | Runs well | Q8_0 | 81.07 GB | 24.8 t/s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Runs great | Q4_K_M | 46.15 GB | 200.2 t/s |
| RTX A6000 NVIDIA | 48 GB | Runs great | Q4_K_M | 46.15 GB | 160.2 t/s |
| Radeon PRO W7900 AMD | 48 GB | Runs great | Q4_K_M | 46.15 GB | 152.8 t/s |
| Apple M1 Max 32GB Apple | 32 GB | Runs great | IQ2_XXS ! | 20.07 GB | 154.4 t/s |
| Apple M2 Max 32GB Apple | 32 GB | Runs great | IQ2_XXS ! | 20.07 GB | 154.4 t/s |
| Jetson AGX Orin 32GB NVIDIA | 32 GB | Runs great | IQ2_XXS ! | 20.1 GB | 85.6 t/s |
| Ryzen AI Max+ 395 32GB AMD | 32 GB | Runs great | IQ2_XXS ! | 20.1 GB | 82.6 t/s |
| Apple M1 Pro 32GB Apple | 32 GB | Runs great | IQ2_XXS ! | 20.07 GB | 77.2 t/s |
| Apple M2 Pro 32GB Apple | 32 GB | Runs great | IQ2_XXS ! | 20.07 GB | 77.2 t/s |
| Apple M5 32GB Apple | 32 GB | Runs great | IQ2_XXS ! | 20.07 GB | 59.1 t/s |
| Apple M4 32GB Apple | 32 GB | Runs great | IQ2_XXS ! | 20.07 GB | 46.3 t/s |
| Intel Core Ultra 9 288V 32GB Intel | 32 GB | Runs great | IQ2_XXS ! | 20.1 GB | 43.9 t/s |
| Ryzen AI 9 HX 370 32GB AMD | 32 GB | Runs great | IQ2_XXS ! | 20.1 GB | 41.3 t/s |
| Apple M4 Max 36GB Apple | 36 GB | Just fits | n/a ! | 29.54 GB | 154.7 t/s |
| Apple M1 Ultra 64GB Apple | 64 GB | Just fits | n/a | 54.43 GB | 133.4 t/s |
Show 35 more devices
| Apple M2 Ultra 64GB Apple | 64 GB | Just fits | n/a | 54.43 GB | 133.4 t/s |
| Apple M4 Max 48GB Apple | 48 GB | Just fits | n/a | 40.8 GB | 117.5 t/s |
| Apple M3 Max 36GB Apple | 36 GB | Just fits | n/a ! | 29.54 GB | 113.4 t/s |
| Apple M4 Pro 24GB Apple | 24 GB | Just fits | n/a ! | 20.07 GB | 105.4 t/s |
| Apple M3 Ultra 96GB Apple | 96 GB | Just fits | n/a | 81.04 GB | 94.9 t/s |
| Apple M4 Max 64GB Apple | 64 GB | Just fits | n/a | 54.43 GB | 91.1 t/s |
| Apple M3 Max 48GB Apple | 48 GB | Just fits | n/a | 40.8 GB | 86.1 t/s |
| Apple M1 Max 64GB Apple | 64 GB | Just fits | n/a | 54.43 GB | 66.7 t/s |
| Apple M2 Max 64GB Apple | 64 GB | Just fits | n/a | 54.43 GB | 66.7 t/s |
| Apple M3 Max 64GB Apple | 64 GB | Just fits | n/a | 54.43 GB | 66.7 t/s |
| Apple M5 24GB Apple | 24 GB | Just fits | n/a ! | 20.07 GB | 59.1 t/s |
| Apple M4 Pro 48GB Apple | 48 GB | Just fits | n/a | 40.8 GB | 58.8 t/s |
| Apple M2 Max 96GB Apple | 96 GB | Just fits | n/a | 81.04 GB | 46.3 t/s |
| Apple M3 Max 96GB Apple | 96 GB | Just fits | n/a | 81.04 GB | 46.3 t/s |
| Apple M4 24GB Apple | 24 GB | Just fits | n/a ! | 20.07 GB | 46.3 t/s |
| Apple M4 Pro 64GB Apple | 64 GB | Just fits | n/a | 54.43 GB | 45.5 t/s |
| Apple M3 Pro 36GB Apple | 36 GB | Just fits | n/a ! | 29.54 GB | 42.5 t/s |
| Ryzen AI 9 HX 370 24GB AMD | 24 GB | Just fits | n/a ! | 20.1 GB | 41.3 t/s |
| Apple M2 24GB Apple | 24 GB | Just fits | n/a ! | 20.07 GB | 38.6 t/s |
| Apple M3 24GB Apple | 24 GB | Just fits | n/a ! | 20.07 GB | 38.6 t/s |
| Jetson AGX Orin 64GB NVIDIA | 64 GB | Just fits | n/a | 54.46 GB | 37 t/s |
| Ryzen AI Max+ 395 64GB AMD | 64 GB | Just fits | n/a | 54.46 GB | 35.7 t/s |
| Ryzen AI Max+ 395 96GB AMD | 96 GB | Just fits | n/a | 81.07 GB | 24.8 t/s |
| GeForce RTX 5090 NVIDIA | 32 GB | Partial offload 48/48 | n/a | 46.15 GB | 84.5 t/s |
| GeForce RTX 4090 NVIDIA | 24 GB | Partial offload 48/48 | n/a | 46.15 GB | 54.6 t/s |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Partial offload 48/48 | n/a | 46.15 GB | 54.6 t/s |
| GeForce RTX 3090 NVIDIA | 24 GB | Partial offload 48/48 | n/a | 46.15 GB | 53.9 t/s |
| GeForce RTX 5090 Laptop NVIDIA | 24 GB | Partial offload 48/48 | n/a | 46.15 GB | 53.4 t/s |
| Radeon RX 7900 XTX AMD | 24 GB | Partial offload 48/48 | n/a | 46.15 GB | 52.4 t/s |
| RTX A5000 NVIDIA | 24 GB | Partial offload 48/48 | n/a | 46.15 GB | 51.7 t/s |
| Radeon RX 7900 XT AMD | 20 GB | Partial offload 48/48 | n/a | 46.15 GB | 44.9 t/s |
| GeForce RTX 5080 NVIDIA | 16 GB | Partial offload 48/48 | n/a | 46.15 GB | 42.9 t/s |
| GeForce RTX 5070 Ti NVIDIA | 16 GB | Partial offload 48/48 | n/a | 46.15 GB | 42.5 t/s |
| GeForce RTX 5080 Laptop NVIDIA | 16 GB | Partial offload 48/48 | n/a | 46.15 GB | 41.5 t/s |
| GeForce RTX 4080 SUPER NVIDIA | 16 GB | Partial offload 48/48 | n/a | 46.15 GB | 41.2 t/s |
How it compares
Other models in the same family or of a similar size, on a 12 GB card and a 24 GB card, each at the best quantization that fits.
| Model | Weights at 4 bits | GeForce RTX 3060 12GB | GeForce RTX 4090 |
|---|---|---|---|
| Qwen3 Next 80B A3B Instruct | 45.7 GB | 32.7 t/s at Q4_K_M | 54.6 t/s at Q4_K_M |
| Qwen3 Coder Next | 44.8 GB | 33.2 t/s at Q4_K_M | 56.4 t/s at Q4_K_M |
| Qwen2.5 VL 72B Instruct | 41.3 GB | does not fit well | does not fit well |
| Qwen2.5 72B Instruct | 40.9 GB | does not fit well | does not fit well |
| GPT OSS 120B | 59.0 GB | 16.9 t/s at MXFP4 | 23.9 t/s at MXFP4 |
| Llama 3.3 70B Instruct | 39.7 GB | does not fit well | does not fit well |
What to buy to run Qwen3 Next 80B A3B Instruct
The cheapest hardware that runs it at a quantization worth using and a speed you would not resent, at an 8k context window.
Cheapest that works
GeForce GTX 1650
$100 used
Q4_K_M · uses 46.15 GB of its 4 GB · about 21.2 tokens/s
Best value
Ryzen AI Max+ 395 64GB
$1,699 whole machine
Q5_K_M · about 35.7 tokens/s · 256 GB/s
| Device | Price | Memory | Runs it at | Speed | |
|---|---|---|---|---|---|
| GeForce GTX 1650 NVIDIA | $100 used | 4 GB | Q4_K_M | 21.2 t/s | Check price |
| GeForce GTX 1660 SUPER NVIDIA | $110 used | 6 GB | Q4_K_M | 29.2 t/s | Check price |
| GeForce GTX 1080 Ti NVIDIA | $150 used | 11 GB | Q4_K_M | 34.3 t/s | Check price |
| GeForce RTX 2060 12GB NVIDIA | $160 used | 12 GB | Q4_K_M | 32.1 t/s | Check price |
| GeForce RTX 2080 Ti NVIDIA | $250 used | 11 GB | Q4_K_M | 35.8 t/s | Check price |
| Arc A770 16GB Intel | $280 used | 16 GB | Q4_K_M | 36.9 t/s | Check price |
| GeForce RTX 3080 12GB NVIDIA | $400 used | 12 GB | Q4_K_M | 38.6 t/s | Check price |
| Radeon RX 7800 XT AMD | $480 | 16 GB | Q4_K_M | 38.7 t/s | Check price |
Indicative prices reviewed 2026-09-01; used prices are marketplace typical. Options that cost more than a cheaper one with no more memory and no more speed are hidden. Price links are Amazon affiliate links. Change the standard, the context window or the budget in the buying tool.
Where it sits in the catalog
Source: Qwen/Qwen3-Next-80B-A3B-Instruct on Hugging Face. Downloaded 281k times in the last month. Published 2025-09-09. Architecture figures are read from the repository's own configuration file, so they move when the model does.
Qwen3 Coder Next
79.67B (3.19B active) · 256k context
Qwen2.5 VL 72B Instruct
73.41B · 125k context
Qwen2.5 72B Instruct
72.71B · 32k context
Qwen3.5 122B A10B
125.09B (8.17B active) · 256k context
Qwen3.5 35B A3B
35.95B (2.9B active) · 256k context
Qwen3 32B
32.76B · 40k context
Qwen2.5 32B Instruct
32.76B · 32k context
Qwen2.5 Coder 32B Instruct
32.76B · 32k context
Questions people ask about Qwen3 Next 80B A3B Instruct
How much VRAM does Qwen3 Next 80B A3B Instruct need?
At Q4_K_M with an 8k context window, Qwen3 Next 80B A3B Instruct needs about 46.15 GB: 45.55 GB of weights, 0.26 GB of KV cache and 0.34 GB of overhead. At the highest-quality file it needs about 80.76 GB.
What is the cheapest GPU that runs Qwen3 Next 80B A3B Instruct?
The smallest consumer device in our catalog that fits it is the Apple M1 Ultra 128GB with 128 GB, running it at Q8_0 at roughly 92.7 tokens per second.
How much does a longer context window add for Qwen3 Next 80B A3B Instruct?
On the memory side, a 32k window adds 0.8 GB of KV cache on top of the weights, and the cache grows more slowly than a conventional model because of its attention design. The longest window the model supports natively is 256k tokens.