2B parameters doesn’t mean 2B-level capability.

MiniCPM5-2B scores 23 on the Artificial Analysis Intelligence Index, ranking #1 among open-source models under 4B parameters.

On the Agentic Index, it scores 20, compared with 2 for LFM2.5-2.6B, Granite 4.2 3B, and Mistral 3 3B.

That’s a 10× higher score, bringing an early form of general-purpose agent capability to the edge.

With a similar token budget:
21K tokens (14K thinking + 7K answer) → MiniCPM5-2B
19K tokens (12K thinking + 7K answer) → Granite 4.2 3B
21K tokens (14K thinking + 7K answer) → LFM2.5-2.6B

2B parameters. Higher intelligence density.