Can SmolLM2 360M Instruct run on the RTX 3070?
Yes. At 8K context the best fit is F16, using about 1.5 GB of the 7.6 GB usable on the RTX 3070. Estimated speed ceiling: ~617 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | Max tok/s (est., 8K) |
|---|---|---|---|
| 2-bit (Q2_K) • | 0.9 GB Fits | 1.0 GB Fits | 2049 |
| 3-bit (Q3_K_M) • | 0.9 GB Fits | 1.0 GB Fits | 1909 |
| 4-bit (Q4_K_M) • | 0.9 GB Fits | 1.1 GB Fits | 1656 |
| 5-bit (Q5_K_M) • | 0.9 GB Fits | 1.1 GB Fits | 1545 |
| 6-bit (Q6_K) • | 1.0 GB Fits | 1.1 GB Fits | 1220 |
| 8-bit (Q8_0) • | 1.0 GB Fits | 1.2 GB Fits | 1159 |
| 16-bit (F16/BF16) • | 1.3 GB Fits | 1.5 GB Fits | 617 |
• real GGUF file size from unsloth/SmolLM2-360M-Instruct-GGUF. Speed is a bandwidth ceiling (448 GB/s), not a benchmark. Methodology.
More
SmolLM2 360M Instruct on every GPU and Mac · Every model that runs on the RTX 3070 · Other models that fit 8 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (HuggingFaceTB/SmolLM2-360M-Instruct), NVIDIA specs. Data as of 2026-11-17.