[KongchangAI]
ModelQwen3 4B

qwen3:4b

A small-to-medium language model in the Qwen3 series with 4B parameters; testing found its tool-calling capability insufficient.

Timeline (last 90 days)

Oct 12

The author embeds an inverted benchmark into the actual system: code deterministically builds and prints a set of boxes, lets the local small model qwen3:4b say one sentence about them, and then the code verifies every reference, citation, and number, discarding any that fail verification and replacing them with plain text written by the code itself.

Unverified50%

All Facts (1)

Source Articles