Can RTX 4070 Ti Super run Magistral Small 1.2 24B?
Tight: GGUF Q4_K_M needs 15.7 GB at 4K, batch 1; headroom 0.3 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 | 15.7 | Tight | Verified file |
| GGUF Q5_K_M | 18.2 | Does not fit | Verified file |
| GGUF Q6_K | 20.9 | Does not fit | Verified file |
| AWQ INT4 | 13.0 | Comfortable | Estimated weights |
| EXL2 4.65bpw | 15.0 | Tight | Estimated weights |
Context capacity · GGUF Q4_K_M
| Context | Total GB | Verdict |
|---|---|---|
| 4K | 15.7 | Tight |
| 16K | 17.7 | Does not fit |
| 32K | 20.5 | Does not fit |
| 128K | 37.0 | Does not fit |
Generation speed
GGUF Q4_K_M: 35–46 tok/s · Derived range