Tag
Recent ternary 1.58-bit LLM releases from small labs demonstrate speed and medical specialization but struggle with long-horizon tasks, with optimism for future models to compete with larger architectures like Qwen.