How much VRAM does Qwen3.6 35B A3B need?
Qwen3.6 35B A3B needs 20.3 GB for its weights at 4-bit, 21.5 GB with an 8K context and 22.0 GB with 32K. At 8-bit, 36.8 GB.
- Parameters
- 36B
- Active per token
- 3B
- Longest context
- 256K
- Released
- Apr 2026
- License
- Apache 2.0
- Strongest at
- Chat, Coding, Agents, Docs
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| 16-bit | 67.0 GB | 68.2 GB | 68.6 GB | 70.5 GB |
| 8-bit | 35.6 GB | 36.8 GB | 37.2 GB | 39.1 GB |
| 6-bit | 27.5 GB | 28.7 GB | 29.1 GB | 31.0 GB |
| 5-bit | 23.8 GB | 25.0 GB | 25.5 GB | 27.4 GB |
| 4-bit | 20.3 GB | 21.5 GB | 22.0 GB | 23.8 GB |
| 3-bit | 16.4 GB | 17.6 GB | 18.0 GB | 19.9 GB |
| 2-bit | 12.6 GB | 13.8 GB | 14.2 GB | 16.1 GB |
| 1.6-bit | 6.9 GB | 8.1 GB | 8.6 GB | 10.5 GB |
One conversation, 16-bit context cache, runtime overhead included.
What runs it
| Machine | Best fit | Context | Speed | First reply |
|---|---|---|---|---|
| RTX 3060 12 GB | Too big | |||
| RTX 4060 Ti 16 GB | Too big | |||
| RTX 3090 | 4-bit | 72K | 190 tok/s | 0.4 s |
| RTX 4090 | 4-bit | 72K | 197 tok/s | 0.2 s |
| RTX 5090 | 6-bit | 96K | 227 tok/s | 0.1 s |
| 2× RTX 3090 | 8-bit | 256K | 136 tok/s | 0.2 s |
| RTX PRO 6000 Blackwell | 8-bit | 256K | 200 tok/s | 0.1 s |
| M4 Max 128 GB | 8-bit | 256K | 68 tok/s | 2.5 s |
| M5 Max 128 GB | 8-bit | 256K | 73 tok/s | 0.7 s |
| Ryzen AI Max+ 395 128 GB | 8-bit | 256K | 40 tok/s | 1.4 s |
| DGX Spark 128 GB | 8-bit | 256K | 52 tok/s | 0.8 s |
| M3 Ultra 256 GB | 8-bit | 256K | 85 tok/s | 1.4 s |
| M5 Ultra 512 GB | 8-bit | 256K | 102 tok/s | 0.4 s |
| H100 SXM | 8-bit | 256K | 264 tok/s | 0.1 s |
| DGX Station GB300 | 8-bit | 256K | 327 tok/s | 0.1 s |
| 8× H200 NVL | 8-bit | 256K | 297 tok/s | 0.1 s |
Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.
How to run it
- Ollama
ollama run qwen3.6:35b-a3b- llama.cpp
llama-server -hf unsloth/Qwen3.6-35B-A3B-GGUF- vLLM
vllm serve Qwen/Qwen3.6-35B-A3B- MLX (Mac)
mlx_lm.chat --model mlx-community/Qwen3.6-35B-A3B-4bit
Questions
How much VRAM does Qwen3.6 35B A3B need?
Qwen3.6 35B A3B needs 20.3 GB for its weights at 4-bit, 21.5 GB with an 8K context and 22.0 GB with 32K. At 8-bit, 36.8 GB.
Can Qwen3.6 35B A3B run on an RTX 4090?
Yes. At 4-bit it holds 72K of context and writes about 197 tokens a second.