Can RTX 5070 run Ministral 3 14B Instruct?
Comfortable: GGUF Q4_K_M needs 9.3 GB at 4K, batch 1; headroom 2.7 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 | 9.3 | Comfortable | Verified file |
| GGUF Q5_K_M | 10.7 | Tight | Verified file |
| GGUF Q8_0 | 15.7 | Does not fit | Verified file |
Context capacity · GGUF Q4_K_M
| Context | Total GB | Verdict |
|---|---|---|
| 4K | 9.3 | Comfortable |
| 16K | 11.4 | Tight |
| 32K | 14.1 | Does not fit |
| 128K | 30.6 | Does not fit |
Generation speed
GGUF Q4_K_M: 59–78 tok/s · Derived range