KimiK2.6 jumps from 50% to 79.9%, Gemini3.0 Flash can even outperform its own Pro—this Meta-System is basically an external big brain for the model, and the coolest part is it doesn't touch the weights at all, relying solely on API external recursive learning. Companies finally don't have to spend money on fine-tuning.

Poetiq's six-member team’s Meta-System set a new high score on LiveCodeBench Pro. This pure API plugin improves itself through recursive self-enhancement to extract task experience, without touching weights or fine-tuning, significantly boosting weak models. After integration, KimiK2.6 increased from 50.0% to 79.9%, Gemini3.0 Flash gained 10 points, surpassing Gemini3.1 Pro, Claude Opus4.7, and GPT5.2 High. GPT5.5 High reached 93.9% through the plugin, Gemini3.1 Pro paired at 90.9%, surpassing Gemini3 Deep Think. Enterprises can improve reasoning capabilities without costly fine-tuning.