Can the Radeon RX 7800 XT run Qwen3.6 27B?
Only with CPU offload at Q4_K_M.
Qwen3.6 27B is 27.8B parameters. At Q4_K_M the weights are about 15.5 GiB; with the 1.5 GiB runtime reserve that is 17.0 GiB against the card's 16GB, about 1.0 GiB more than the card has.
All three quants
16GB card| Quant | Weights | With reserve | Verdict | Headroom |
|---|---|---|---|---|
| Q4_K_Mthe everyday quant | 15.5 GiB | 17.0 GiB | Only with CPU offload | short 1.0 GiB |
| Q8_0near-lossless | 27.4 GiB | 28.9 GiB | Only with CPU offload | short 12.9 GiB |
| FP16full weights | 51.8 GiB | 53.3 GiB | Does not fit | short 37.3 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 15.0 GiB
- GeForce RTX 3090 (used) 24GB · headroom 7.0 GiB
- GeForce RTX 4090 (used) 24GB · headroom 7.0 GiB
- Radeon RX 7900 XTX 24GB · headroom 7.0 GiB
What the Radeon RX 7800 XT does run well → · Qwen3.6 27B on every card →
About the model
Qwen3.6 27B by Alibaba Qwen: 262,144-token native context, text + vision, released 2026 under the Apache-2.0. Model card → · Hosted pricing on PicksByModel →