Can RTX 5090 run Qwen3.5 9B?
Comfortable: GGUF Q8_0 needs 10.4 GB at 4K, batch 1; headroom 21.6 GB.
Text input, FP16 KV cache, one card. Unified-memory capacity uses the GPU guide’s usable limit; negative headroom may require host offload. Speed source settings may differ from this 4K estimate.
Measured runs
No measured run for this pair. Sizing is an estimate, not a runtime test.
Quant memory at 4K
| Quant | Total GB | Verdict | Weight sizing |
|---|---|---|---|
| GGUF Q4_K_M | 6.6 | Comfortable | Verified file |
| GGUF Q8_0 | 10.4 | Comfortable | Verified file |
| GGUF Q5_K_M | 7.6 | Comfortable | Verified file |
| GGUF Q6_K | 8.5 | Comfortable | Verified file |
Context capacity · GGUF Q8_0
| Context | Total GB | Verdict |
|---|---|---|
| 4K | 10.4 | Comfortable |
| 16K | 10.8 | Comfortable |
| 32K | 11.3 | Comfortable |
| 128K | 14.7 | Comfortable |
Generation speed
GGUF Q8_0: 112–148 tok/s · Derived range