MicroLLM lab — tiny LLMs, Q4, in your browser

Global World Wed, 30 Sep 2026 18:59:40 GMT Moderate confidence — 64/100

max new Next action: Objective checks (regex / exact tokens), not writing quality.
A 135M model is allowed to fail — that is the measurement.

Unverified

max new Next action: Objective checks (regex / exact tokens), not writing quality.
A 135M model is allowed to fail — that is the measurement.

Sources: Stateofutopia