LFM2.5 1.2B Thinking
Liquid AI
Multilingual Thinking Tool Calls
LFM2.5 1.2B Thinking is a 1.17-billion-parameter hybrid convolution-attention model from Liquid AI, optimized for on-device chain-of-thought reasoning. It produces thinking traces before answering, delivering math and logic performance that rivals models with 40% more parameters. The model supports tool calling and eight languages including English, French, German, and Spanish. With a 128K context window and flash attention, it fits under 1 GB as a Q4 GGUF for efficient edge deployment on mobile and consumer hardware.
Hardware Configuration
Optional — for precise deployment recommendations
| Quantization | Quality | Size | Fit |
|---|---|---|---|
| FP16 | Full precision | 2.18 GB | — |
| Q8_0 | High | 1.16 GB | — |
| Q6_K | High | 0.9 GB | — |
| Q5_K_M | Medium | 0.79 GB | — |
| Q4_K_M | Medium | 0.68 GB | — |
| Q4_0 | Medium | 0.65 GB | — |
Last updated: March 5, 2026