MiniCPM5
A series of edge-side small language models developed by ModelBest (面壁智能), designed for deployment on edge devices. Built on the standard LLaMA architecture, the series supports long-context processing and tool calling, achieving strong capabilities with a minimal parameter count.
Timeline (last 90 days)
Latent Compaction achieves retrieval gains on both Qwen3 and MiniCPM5, demonstrating cross-model-family transferability.
In the summarization task, 4 out of the 8 models failed: the 4-billion-parameter Qwen3.5 fabricated content, LFM2.5 altered facts, Bonsai added nonexistent information, and MiniCPM5 missed key points.
All Facts (2)
Latent Compaction achieves retrieval gains on both Qwen3 and MiniCPM5, demonstrating cross-model-family transferability.
50%UnverifiedIn the summarization task, 4 out of the 8 models failed: the 4-billion-parameter Qwen3.5 fabricated content, LFM2.5 altered facts, Bonsai added nonexistent information, and MiniCPM5 missed key points.
50%