RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can SmolVLM 500M Instruct run on the RTX 3090?

Yes. At 8K context the best fit is F16, using about 1.6 GB of the 22.8 GB usable on the RTX 3090. Estimated speed ceiling: ~1141 tokens/sec.

Memory by quant and context

Quant4K ctx8K ctxMax tok/s (est., 8K)
2-bit (Q2_K)0.8 GB Fits1.0 GB Fits4918
3-bit (Q3_K_M)0.9 GB Fits1.0 GB Fits3774
4-bit (Q4_K_M)0.9 GB Fits1.1 GB Fits3017
5-bit (Q5_K_M)1.0 GB Fits1.1 GB Fits2589
6-bit (Q6_K)1.0 GB Fits1.2 GB Fits2249
8-bit (Q8_0) •1.1 GB Fits1.2 GB Fits2143
16-bit (F16/BF16) •1.4 GB Fits1.6 GB Fits1141

• real GGUF file size from ggml-org/SmolVLM-500M-Instruct-GGUF. Speed is a bandwidth ceiling (936 GB/s), not a benchmark. Methodology.

More

SmolVLM 500M Instruct 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 (HuggingFaceTB/SmolVLM-500M-Instruct), NVIDIA specs. Data as of 2026-11-17.