Can Qwen3.5 35B A3B run on the RTX 3090?
Yes. At 8K context the best fit is Q3_K_M, using about 16.3 GB of the 22.8 GB usable on the RTX 3090. Estimated speed ceiling: ~686 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | 13.5 GB Fits | 13.6 GB Fits | 14.1 GB Fits | 15.9 GB Fits | 832 |
| 3-bit (Q3_K_M) • | 16.2 GB Fits | 16.3 GB Fits | 16.7 GB Fits | 18.6 GB Fits | 686 |
| 4-bit (Q4_K_M) • | 21.5 GB Tight | 21.5 GB Tight | 22.0 GB Tight | 23.9 GB No | 509 |
| 5-bit (Q5_K_M) • | 25.4 GB No | 25.5 GB No | 25.9 GB No | 27.8 GB No | — |
| 6-bit (Q6_K) • | 27.8 GB No | 27.9 GB No | 28.4 GB No | 30.2 GB No | — |
| 8-bit (Q8_0) • | 35.3 GB No | 35.4 GB No | 35.9 GB No | 37.7 GB No | — |
| 16-bit (F16/BF16) • | 65.6 GB No | 65.6 GB No | 66.1 GB No | 68.0 GB No | — |
• real GGUF file size from unsloth/Qwen3.5-35B-A3B-GGUF. Speed is a bandwidth ceiling (936 GB/s), not a benchmark. Methodology.
More
Qwen3.5 35B A3B on every GPU and Mac · Every model that runs on the RTX 3090 · Other models that fit 24 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (Qwen/Qwen3.5-35B-A3B), NVIDIA specs. Data as of 2026-11-17.