Can the GeForce RTX 5080 run Qwen3.6 35B-A3B?
Only with CPU offload at Q4_K_M.
Qwen3.6 35B-A3B is 36B parameters (MoE; all experts resident). At Q4_K_M the weights are about 20.1 GiB; with the 1.5 GiB runtime reserve that is 21.6 GiB against the card's 16GB, about 5.6 GiB more than the card has.
All three quants
16GB card| Quant | Weights | With reserve | Verdict | Headroom |
|---|---|---|---|---|
| Q4_K_Mthe everyday quant | 20.1 GiB | 21.6 GiB | Only with CPU offload | short 5.6 GiB |
| Q8_0near-lossless | 35.5 GiB | 37.0 GiB | Does not fit | short 21.0 GiB |
| FP16full weights | 67.1 GiB | 68.6 GiB | Does not fit | short 52.6 GiB |
Verdict rules: fits comfortably when weights plus reserve sit within 85% of VRAM; tight when they fit with little left for context; CPU offload when they exceed VRAM by up to 2x (runs, slowly, with layers in system RAM); no beyond that. The KV cache per token depends on the model's architecture and is not modelled here.
Not this card. These fit it comfortably at Q4.
- GeForce RTX 5090 32GB · headroom 10.4 GiB
What the GeForce RTX 5080 does run well → · Qwen3.6 35B-A3B on every card →
About the model
Qwen3.6 35B-A3B by Alibaba Qwen: 262,144-token native context, text + vision, released 2026 under the Apache-2.0. Model card → · Hosted pricing on PicksByModel →