Qwen3 Next Moe — Hardware Requirements & GPU Compatibility
ChatFunctions
Qwen3 Next Moe is a 3M-parameter open language model from tiny-random in the Qwen 3 family. It supports a context window of up to 262,144 tokens. At Q4_K_M it needs about 0.30 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 256K)Best on 24 GB (Q4_K_M · ~2184 tok/s · 256K)Macs from 8 GB (Q4_K_M · ~239 tok/s)
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Qwen3 Next Moe
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
vllm serve tiny-random/qwen3-next-moe --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on RTX 3080 Ti (12 GB) (11 GB headroom) · Q4_K_M
- Generation speed
- ~1977 tok/s
- generation speed
- Cost per 1M output tokens
- $0.017
- per 1M output tokens
How Much VRAM Does Qwen3 Next Moe Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 0.3 GB | 0.3 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 256K context · ~2184 tok/s | |
| 0.3 GB | 0.3 GB | 3-bit medium — 3-bit medium quantization | Fits: 256K context · ~2184 tok/s | |
| 0.3 GB | 0.3 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 256K context · ~2184 tok/s | |
| 0.3 GB | 0.3 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 256K context · ~2184 tok/s | |
| 0.3 GB | 0.3 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 256K context · ~2184 tok/s | |
| 0.3 GB | 0.3 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 256K context · ~2184 tok/s | |
| 0.3 GB | 0.3 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 256K context · ~2114 tok/s |
Selected Q4_K_M
4-bit medium quantization0.3GBat 16K context
- Weights
- 0.1 GB
- KV cache
- 0.0 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 23.7 GB free · ~2184 tok/s
≥50 tok/s needs ≥23 GB/s · RTX 5090 ~3883, RTX 3090 Ti ~2184
Which Hardware Can Run Qwen3 Next Moe?
Q4_K_M · 0.3 GBQwen3 Next Moe (Q4_K_M) requires 0.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 1+ GB is recommended. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 163 devices with unified memory can run Qwen3 Next Moe, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB).
≤ 11 GB
≤ 11 GB tier →fits · 7.7 GB free · up to 256K ctx
16 GPUs · 1 Mac class · 2 device families
RTX 3080~1647Simulate output on RTX 3080RTX 2080 Ti~1335Simulate output on RTX 2080 TiRTX 3070 Ti~1318Simulate output on RTX 3070 TiM3*8 GB~239Simulate output on iMac 24" M3 (8 GB)Orin Nano~221Simulate output on NVIDIA Jetson Orin Nano 8GB (Super)iPhone~159Simulate output on iPhone 1720 GB
fits · 19.7 GB free · up to 256K ctx
1 GPU
RX 7900 XT~1600Simulate output on RX 7900 XT32–40 GB
fits · 31.7 GB free · up to 256K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~3883Simulate output on RTX 5090M5 Max36 GB~1073Simulate output on Mac Studio M5 Max (36 GB)M532 GB~358Simulate output on MacBook Air 13" M5 (32 GB)Orin~444Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~293Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to Qwen3 Next Moe
Qwen 3 family · 79 models →Specifications
- Publisher
- tiny-random
- Family
- Qwen 3
- Parameters
- 3M
- Architecture
- Qwen3NextForCausalLM
- Context Length
- 262,144 tokens
- Vocabulary Size
- 151,936
- Release Date
- 2025-09-12
Frequently Asked Questions
How much VRAM does Qwen3 Next Moe need?
Qwen3 Next Moe requires 0.3 GB of VRAM at Q4_K_M, or 0.3 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 3M × 4.8 bits ÷ 8 = 0 GB
KV Cache + Overhead ≈ 0.3 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 0.3 GB (at full 262K context)
VRAM usage by quantization
What's the best quantization for Qwen3 Next Moe?
For Qwen3 Next Moe, Q4_K_M (0.3 GB) offers the best balance of quality and VRAM usage. Q5_K_M (0.3 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 0.3 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Qwen3 Next Moe on a Mac?
Yes — MacBook Air 13" M3 (8 GB) and 145 other Macs can run Qwen3 Next Moe. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 0.3 GB of usable unified memory (RAM minus macOS overhead).
Can I run Qwen3 Next Moe locally?
Yes — Qwen3 Next Moe can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Qwen3 Next Moe?
At Q4_K_M, Qwen3 Next Moe can reach ~16000 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~2184 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 0.3 × 0.65 = ~17333 tok/s
Estimated speed at Q4_K_M (0.3 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Qwen3 Next Moe?
At Q4_K_M, the download is about 0.01 GB. The full-precision BF16 version is 0.01 GB. The smallest option (Q2_K) is 0.01 GB.
Which GPUs can run Qwen3 Next Moe?
52 consumer GPUs can run Qwen3 Next Moe at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 52 GPUs have plenty of headroom for comfortable inference.
Which devices can run Qwen3 Next Moe?
163 devices with unified memory can run Qwen3 Next Moe at Q4_K_M (0.3 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.