Can SingVL S1 4B run on the RTX 3090?
Yes. At 8K context the best fit is F16, using about 9.0 GB of the 22.8 GB usable on the RTX 3090. Estimated speed ceiling: ~111 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 2.8 GB Fits | 2.9 GB Fits | 3.7 GB Fits | 6.7 GB Fits | 489 |
| 3-bit (Q3_K_M) • | 3.1 GB Fits | 3.2 GB Fits | 4.0 GB Fits | 7.0 GB Fits | 414 |
| 4-bit (Q4_K_M) • | 3.5 GB Fits | 3.6 GB Fits | 4.4 GB Fits | 7.4 GB Fits | 346 |
| 5-bit (Q5_K_M) • | 3.9 GB Fits | 4.0 GB Fits | 4.7 GB Fits | 7.7 GB Fits | 304 |
| 6-bit (Q6_K) • | 4.2 GB Fits | 4.3 GB Fits | 5.1 GB Fits | 8.1 GB Fits | 270 |
| 8-bit (Q8_0) • | 5.2 GB Fits | 5.3 GB Fits | 6.0 GB Fits | 9.0 GB Fits | 209 |
| 16-bit (F16/BF16) • | 8.8 GB Fits | 9.0 GB Fits | 9.7 GB Fits | 12.7 GB Fits | 111 |
• real GGUF file size from maternion/SingVL-S1-4B-GGUF. Speed is a bandwidth ceiling (936 GB/s), not a benchmark. Methodology.
More
SingVL S1 4B on every GPU and Mac · Every model that runs on the RTX 3090 · Other models that fit 24 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (inclusionAI/SingVL_S1_4B), NVIDIA specs. Data as of 2026-11-17.