Ollama launches LFM3 and d1 for local agents
The new LFM3 model (1.7B parameters) and d1 (2B parameters) run on Apple Silicon, supporting tool calling and audio processing.
- Field
- training language models
- What they did
- A new family of compact language models named LFM3 has been created using a tri-hybrid architecture, featuring versions for text, images, audio, and multimodal tasks.
- Why it matters
- These models are designed for local agents and tool calling, with specific support for Apple hardware.
#ollama#llm#local agents#apple silicon#tool calling



