Can Qwen3.5 35B A3B run on the RX 7900 XT?
Yes. At 8K context the best fit is Q3_K_M, using about 17.0 GB of the 19.0 GB usable on the RX 7900 XT. Estimated speed ceiling: ~560 tokens/sec.
Better quality with offload: 4-bit (Q4_K_M)
Runs with offload 4-bit (Q4_K_M) at 8K context: all attention layers plus the experts of 32 of 40 MoE layers on the RX 7900 XT (about 17.2 GB), and the experts of the other 8 layers in system RAM (about 3.7 GB). Needs a PC with at least 32 GB of system RAM. Estimated speed: 96–131 tokens/sec (usable for chat), versus a bandwidth ceiling of ~430 tok/s if it fit fully in VRAM.
llama-server -hf bartowski/Qwen_Qwen3.5-35B-A3B-GGUF:Q4_K_M -c 8192 -ngl 999 --n-cpu-moe 8 --no-mmproj
Estimate, not a benchmark. The command downloads the GGUF from Hugging Face. The speed range spans dual-channel DDR4-3200 to DDR5-6000 system RAM; your CPU, RAM speed and llama.cpp version move it. If llama.cpp runs out of GPU memory, raise --n-cpu-moe by 1-2; if VRAM is left over, lower it for more speed. Recent llama.cpp builds can also pick the split automatically. How we estimate offload.
+ 0.8 GB with image input. The figures on this page are for text only. With image input enabled, the vision projector (mmproj) file loads alongside the model: Q3_K_M at 8K context then needs about 17.8 GB (tight). llama.cpp's -hf downloads and loads it automatically unless you add --no-mmproj. Calculator with image input on.
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 | 16.0 GB Fits | 710 |
| 3-bit (Q3_K_M) • | 16.9 GB Fits | 17.0 GB Fits | 17.4 GB Tight | 19.3 GB No | 560 |
| 4-bit (Q4_K_M) • | 21.7 GB No | 21.8 GB No | 22.3 GB No | 24.1 GB No | — |
| 5-bit (Q5_K_M) • | 25.1 GB No | 25.2 GB No | 25.6 GB No | 27.5 GB No | — |
| 6-bit (Q6_K) • | 29.8 GB No | 29.9 GB No | 30.3 GB No | 32.2 GB No | — |
| 8-bit (Q8_0) • | 36.2 GB No | 36.2 GB No | 36.7 GB No | 38.6 GB No | — |
| 16-bit (F16/BF16) • | 67.1 GB No | 67.2 GB No | 67.7 GB No | 69.6 GB No | — |
• real GGUF file size from bartowski/Qwen_Qwen3.5-35B-A3B-GGUF. Speed is a bandwidth ceiling (800 GB/s), not a benchmark. Methodology.
What fits instead
- Cheapest listed GPU that runs Qwen3.5 35B A3B: RTX 5090 (32 GB, launch MSRP $1,999) at Q5_K_M, fully in VRAM at 8K context.
- Largest popular model that fits the RX 7900 XT: MiniMax H3 Image PE (27B) at Q4_K_M, about 16.8 GB at 8K context.
"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.
More
Qwen3.5 35B A3B on every GPU and Mac · Every model that runs on the RX 7900 XT · Other models that fit 24 GB VRAM · Try other settings in the calculator
Guides: Best GPU for local LLMs by budget · Can I run gpt-oss, Llama 70B or Qwen3? · Mac vs PC for local LLMs
Data: Hugging Face Hub (Qwen/Qwen3.5-35B-A3B), AMD specs. Model data last checked 2026-11-18; refreshed hourly.