ZelligeAI/Tessera-27B-B-CJK-RL: A CJK-thinking coding-agent model with reinforcement learning

Tessera-27B-B-CJK-RL is a coding-agent model built on Qwen3.8-27B that thinks in a compressed CJK shorthand. Developed with eight updates of reinforcement learning on math and code, the beta release improves performance on competition math while retaining most of the base model's per-turn thinking savings. Open article →