Community converts Qwen3-0.6B into MoE model

A developer converted Qwen3-0.6B into a Mixture-of-Experts model and released GGUF versions. Lacking Qwen3's shared expert, quality dropped; a pseudo-shared expert recovered some of it. MoE perplexity and task scores were worse than dense, but generation speed rose sharply.

• ProgramAsWeights Compiles English into Local Neural Programs2026-09-20
• Qwen3-0.6B drives real desktop Chrome on old phone2026-09-09