RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can LFM2.5 350M run on the 2x RTX 3090?

Yes. At 8K context the best fit is F16, using about 1.9 GB of the 45.6 GB usable on the 2x RTX 3090. Estimated speed ceiling: ~658 tokens/sec.

Memory by quant and context

Quant4K ctx8K ctx32K ctxMax tok/s (est., 8K)
2-bit (Q2_K)0.7 GB Fits0.7 GB Fits1.0 GB Fits7041
3-bit (Q3_K_M)0.7 GB Fits0.8 GB Fits1.1 GB Fits5402
4-bit (Q4_K_M) •0.8 GB Fits0.8 GB Fits1.1 GB Fits4082
5-bit (Q5_K_M) •0.8 GB Fits0.9 GB Fits1.1 GB Fits3595
6-bit (Q6_K) •0.8 GB Fits0.9 GB Fits1.2 GB Fits3190
8-bit (Q8_0) •0.9 GB Fits1.0 GB Fits1.3 GB Fits2468
16-bit (F16/BF16) •1.9 GB Fits1.9 GB Fits2.2 GB Fits658

• real GGUF file size from LiquidAI/LFM2.5-350M-GGUF. Speed is a bandwidth ceiling (936 GB/s), not a benchmark. Methodology.

More

LFM2.5 350M on every GPU and Mac · Every model that runs on the 2x RTX 3090 · Try other settings in the calculator

Data: Hugging Face Hub (LiquidAI/LFM2.5-350M), NVIDIA specs. Data as of 2026-11-17.