LFM2.5 Models — Hardware Requirements
14 LFM2.5 models from Liquid AI and the community, from the smallest that runs in 0.6 GB of VRAM at Q4 up to 24.0B parameters. Every row links to full quantization tables and GPU compatibility.
All LFM2.5 Models by Size
| Model | Params | Runs from (Q4) | Context | Publisher | Quant downloads |
|---|---|---|---|---|---|
| LFM2.5 230M | 230M | 0.6 GB | 128K | ||
| LFM2.5 2.6B DSpark | 328M | 0.7 GB | 128K | ||
| LFM2.5 350M | 354M | 0.8 GB | 128K | ||
| LFM2.5 350M Base | 354M | 0.8 GB | 128K | ||
| LFM2.5 VL 450M | 449M | 0.8 GB | 128K | ||
| LFM2.5 1.2B Instruct | 1.2B | 1.2 GB | 128K | ||
| LFM2.5 1.2B Thinking | 1.2B | 1.2 GB | 128K | ||
| LFM2.5 1.2B Base | 1.2B | 1.2 GB | 128K | ||
| LFM2.5 2.6B | 2.7B | 2.3 GB | 131K | ||
| LFM2.5 8B A1B | 8.5B | 5.6 GB | 128K | ||
| LFM2.5 24B A2B | 24.0B | 15.0 GB | 131K |
Community variants and fine-tunes (9)
| Model | Params | Runs from (Q4) | Context | Publisher | Quant downloads |
|---|---|---|---|---|---|
| LFM2.5 230M ONNX | 230M | 0.6 GB | 128K | ||
| LFM2.5 230M Chess | 231M | 0.6 GB | 128K | ||
| LFM2.5 1.2B JP 202606 | 1.2B | 1.2 GB | 128K | ||
| LFM2.5 1.2B Instruct Uncensored | 1.2B | 1.2 GB | 128K | ||
| Lfm2.5 2.6B Fable5 Coding Agent Heretic | 2.7B | 2.3 GB | 131K | ||
| LFM2.5 8B A1B DSpark | 8B | 5.3 GB | 128K | ||
| LFM2.5 8B A1B Base | 8.5B | 5.6 GB | 128K | ||
| LFM2.5 8B A1B Opus Distil | 8.5B | 5.6 GB | 128K | ||
| Huihui LFM2.5 8B A1B Abliterated | 8.5B | 5.6 GB | 128K |
Frequently Asked Questions
- How much VRAM do I need to run a LFM2.5 model?
- The smallest LFM2.5 model, LFM2.5 230M, runs from 0.6 GB of VRAM at Q4_K_M (smaller low-bit quantizations start at 0.4 GB with a noticeable quality loss). Larger family members need proportionally more — see the table above for every model.
- Which LFM2.5 models can I run on a 16 GB GPU?
- 20 of 20 LFM2.5 models fit in 16 GB of VRAM at Q4_K_M, including LFM2.5 2.6B, LFM2.5 1.2B Instruct, LFM2.5 230M.
- What is the most popular LFM2.5 model to run locally?
- LFM2.5 2.6B is the most downloaded LFM2.5 model in local-friendly quantized formats. It runs from 2.3 GB of VRAM at Q4_K_M.