Qwen · 35B · MoE
Can you run Qwen3.6-35B-A3B locally?
Qwen3.6-35B-A3B is a mixture-of-experts model with 35B parameters, 3B active per token. At 4-bit or better it runs on 59 of the 120 hardware configurations listed here; the smallest is the RTX 5090 (32 GB). On a 128 GB DGX Spark it runs at Q8_0, at roughly 45 tokens per second (our est.).
Data as of 2 November 2026
Fastest machines for Qwen3.6-35B-A3B
Every machine that runs it, fastest first. Filter by type; your machine is marked. Compare machines running it side by side
- 1RTX 5090Graphics card · 32 GB~346 tok/s ModelFit est.Q4_K_M · 24 of 31 GB
- 2RTX PRO 6000 BlackwellGraphics card · 96 GB~251 tok/s ModelFit est.Q8_0 · 41 of 95 GB
- 32x RTX 5090Graphics card · 64 GB~234 tok/s ModelFit est.Q8_0 · 41 of 62 GB
- 42x RTX PRO 6000 BlackwellGraphics card · 192 GB~234 tok/s ModelFit est.Q8_0 · 41 of 190 GB
- 5M5 Max · 48 GB · 40cMac · 48 GB~163 tok/s ModelFit est.Q4_K_M · 24 of 36 GB
- 6M5 UltraMac · 96–512 GB · 5 variants~160 tok/s ModelFit est.Q8_0 · 41 GB
- 72x RTX 4090Graphics card · 48 GB~154 tok/s ModelFit est.Q8_0 · 41 of 46 GB
- 82x RTX 3090 TiGraphics card · 48 GB~143 tok/s ModelFit est.Q8_0 · 41 of 46 GB
- 9M4 Max · 48 GB · 40cMac · 48 GB~137 tok/s ModelFit est.Q4_K_M · 24 of 36 GB
- 102x RTX 3090Graphics card · 48 GB~136 tok/s ModelFit est.Q8_0 · 41 of 46 GB
- 11M5 Max · 36 GB · 32cMac · 36 GB~128 tok/s ModelFit est.Q4_K_M · 24 of 24 GB
- 122x RX 7900 XTXGraphics card · 48 GB~117 tok/s ModelFit est.Q8_0 · 41 of 46 GB
- 13M3 UltraMac · 96–512 GB · 5 variants~112 tok/s ModelFit est.Q8_0 · 41 GB
- 14M3 Max · 48 GB · 40cMac · 48 GB~109 tok/s ModelFit est.Q4_K_M · 24 of 36 GB
- 15M4 Max · 36 GB · 32cMac · 36 GB~108 tok/s ModelFit est.Q4_K_M · 24 of 24 GB
- 16M5 MaxMac · 64–128 GB · 2 variants~103 tok/s ModelFit est.Q8_0 · 41 GB
- 17M2 UltraMac · 64–192 GB · 7 variants~99 tok/s ModelFit est.Q8_0 · 41 GB
- 18M5 ProMac · 48 GB · 2 variants~99 tok/s ModelFit est.Q4_K_M · 24 GB
- 19M1 UltraMac · 64–128 GB · 4 variants~88 tok/s ModelFit est.Q8_0 · 41 GB
- 20M4 MaxMac · 64–128 GB · 2 variants~87 tok/s ModelFit est.Q8_0 · 41 GB
- 21M3 Max · 36 GB · 30cMac · 36 GB~85 tok/s ModelFit est.Q4_K_M · 24 of 24 GB
- 22M4 ProMac · 48 GB · 2 variants~83 tok/s ModelFit est.Q4_K_M · 24 GB
- 23M3 MaxMac · 64–128 GB · 2 variants~68 tok/s ModelFit est.Q8_0 · 41 GB
- 24M2 MaxMac · 64–96 GB · 3 variants~64 tok/s ModelFit est.Q8_0 · 41 GB
- 25M5 ProMac · 64 GB · 2 variants~60 tok/s ModelFit est.Q8_0 · 41 GB
- 26M1 MaxMac · 64 GB · 2 variants~57 tok/s ModelFit est.Q8_0 · 41 GB
- 27M3 Max · 96 GB · 30cMac · 96 GB~52 tok/s ModelFit est.Q8_0 · 41 of 72 GB
- 28M4 ProMac · 64 GB · 2 variants~50 tok/s ModelFit est.Q8_0 · 41 GB
- 29M3 Pro · 36 GBMac · 36 GB~49 tok/s ModelFit est.Q4_K_M · 24 of 24 GB
- 30DGX Spark (GB10)AI mini PC · 128 GB~45 tok/s our est.Q8_0 · 41 of 122 GB
- 31Ryzen AI Max+ 395 · 64 GBAI mini PC · 64 GB~42 tok/s our est.Q8_0 · 41 of 48 GB
- 32Ryzen AI Max+ 395 · 128 GBAI mini PC · 128 GB~40 tok/s ModelFit est.Q8_0 · 41 of 96 GB
Too big for 61 machines
RTX 4090, RTX 3090 Ti, RTX 3090, RX 7900 XTX, M2 Max · 32 GB · 38c, M2 Max · 32 GB · 30c, M1 Max · 32 GB · 32c, M1 Max · 32 GB · 24c, M2 Pro · 32 GB, M1 Pro · 32 GB, M6 · 32 GB · 12c, M5 · 32 GB · 10c, M5 · 32 GB · 8c, M4 · 32 GB, RX 7900 XT, M5 Pro · 24 GB · 20c, M5 Pro · 24 GB · 16c, M4 Pro · 24 GB · 20c, M4 Pro · 24 GB · 16c, M6 · 24 GB · 12c, M6 · 24 GB · 10c, M5 · 24 GB · 10c, M5 · 24 GB · 8c, M4 · 24 GB, M3 · 24 GB, M2 · 24 GB, RTX 5080, RTX 5070 Ti, RTX 5060 Ti, RTX 4080 SUPER, RTX 4080, RTX 4070 Ti SUPER, RTX 4060 Ti, Tesla T4, Tesla P100, RX 9070 XT, RX 9060 XT, M3 Pro · 18 GB, RTX 5070, RTX 4070 Ti, RTX 4070 SUPER, RTX 4070, RTX 3060, M2 Pro · 16 GB, M1 Pro · 16 GB, M6 · 16 GB · 12c, M6 · 16 GB · 10c, M5 · 16 GB · 10c, M5 · 16 GB · 8c, M4 · 16 GB, M3 · 16 GB, M2 · 16 GB, M1 · 16 GB, RTX 5060, RTX 4060, RTX 3070, RTX 3060 Ti, RTX 2070, M3 · 8 GB, M2 · 8 GB, M1 · 8 GB. Qwen3.6-35B-A3B needs about 24 GB at its smallest build.
Speeds are single-user decode estimates, not measurements. "ModelFit est." comes from ModelFit's model fitted on published runs (CC BY 4.0); "our est." is memory bandwidth divided by bytes read per token, times 0.55. Memory is what the best build needs, against what the machine can give a model. Method
DGX Spark setup notes: GPUwerk guide for Qwen3.6-35B-A3B.
About the model
- Parameters
- 35B
- Active per token
- 3B
- Type
- Mixture of experts
- Good for
- Fast mid-size MoE, long context
- Model card
- Hugging Face
- Hugging Face count
- 35.95B (agrees)
Builds and how to run them
| Quantization | Load | Min. RAM | Command |
|---|---|---|---|
| Q4_K_M | 22 GB | 32 GB | ollama run qwen3.6:35b-a3b |
| Q8_0 | 39 GB | 64 GB | ollama run qwen3.6:35b-a3b-q8_0 |
Load, minimum RAM and Ollama commands: ModelFit, CC BY 4.0. Load is memory with the model loaded, before long contexts.
Questions
Can I run Qwen3.6-35B-A3B on a Mac?
Yes on a Mac with enough memory: the smallest Apple configuration listed here that runs it at 4-bit or better is the M5 Max · 36 GB · 32c.
How much memory does Qwen3.6-35B-A3B need?
About 24 GB at the smallest build in our data, including a margin for the runtime and an 8K-token context. Longer contexts need more.
Does Qwen3.6-35B-A3B run on a DGX Spark?
Yes, at Q8_0. The Spark has 128 GB of unified memory, about 121.6 GB usable.