mohommed/LFM3-1.7B-Frontier-SmartAgent-GGUF
GGUF quantizations of mohommed/LFM3-1.7B-Frontier-SmartAgent with 512K NTK-YaRN Context and 4-Stage Frontier <think> Reasoning.
Quantization Guide
For a compact 1.7B Tri-Hybrid architecture (shortconv + attention), Q8_0 (1.84 GB) or Q6_K (1.42 GB) is strongly recommended to retain 99.8% of full-precision coding and tool-calling reliability.
| File | Quantization | Approx. Size | Recommendation |
|---|---|---|---|
lfm3-1.7b-frontier-f16.gguf |
F16 | 3.45 GiB | Full-precision reference |
lfm3-1.7b-frontier-q8_0.gguf |
Q8_0 | 1.84 GiB | Recommended for Agentic & Coding Use |
lfm3-1.7b-frontier-q6_k.gguf |
Q6_K | 1.42 GiB | High quality, low VRAM |
lfm3-1.7b-frontier-q5_k_m.gguf |
Q5_K_M | 1.22 GiB | Balanced speed & quality |
lfm3-1.7b-frontier-q4_k_m.gguf |
Q4_K_M | 1.02 GiB | Fast CPU / laptop inference |
lfm3-1.7b-frontier-q3_k_m.gguf |
Q3_K_M | 0.82 GiB | Ultra-compact edge devices |
Run with Ollama
ollama run hf.co/mohommed/LFM3-1.7B-Frontier-SmartAgent-GGUF:Q8_0