Can LFM2.5 24B A2B run on the RTX 3070?
No. Even at 2-bit with 8K context, LFM2.5 24B A2B needs about 9.2 GB, more than the 7.6 GB usable on the RTX 3070.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | 9.1 GB No | 9.2 GB No | 9.6 GB No | 11.5 GB No | — |
| 3-bit (Q3_K_M) | 11.6 GB No | 11.7 GB No | 12.2 GB No | 14.1 GB No | — |
| 4-bit (Q4_K_M) • | 14.3 GB No | 14.3 GB No | 14.8 GB No | 16.7 GB No | — |
| 5-bit (Q5_K_M) • | 16.6 GB No | 16.7 GB No | 17.1 GB No | 19.0 GB No | — |
| 6-bit (Q6_K) • | 19.1 GB No | 19.1 GB No | 19.6 GB No | 21.5 GB No | — |
| 8-bit (Q8_0) • | 24.5 GB No | 24.5 GB No | 25.0 GB No | 26.9 GB No | — |
| 16-bit (F16/BF16) • | 90.1 GB No | 90.1 GB No | 90.6 GB No | 92.5 GB No | — |
• real GGUF file size from LiquidAI/LFM2.5-24B-A2B-GGUF. Speed is a bandwidth ceiling (448 GB/s), not a benchmark. Methodology.
Hardware that does fit it
RTX 3080 10GBRTX 3060 12GBRTX 3080 12GBRTX 4070RTX 4070 SuperRTX 4070 Ti
More
LFM2.5 24B A2B 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 (LiquidAI/LFM2.5-24B-A2B), NVIDIA specs. Data as of 2026-11-17.