Wësche(@WescheNex1q):Le Chonk vs full GLM-5.3 on the same voxel pagoda prompt. -50K tokens vs 126K. 19.5 min vs 72. Both ran first try, untouched. Thinking on, temp 0.7, no output cap. GLM local on 4 Sparks, Mistral via @OpenRouterAI I think it did a pretty impressive job

Wësche

@WescheNex1q

Day time artist and night time AI enthusiast. Building & benchmarking frontier LLMs on 4x DGX Spark clusters + Mac. Creator of Vesica Studio. Houston

加入 January 2013

731 正在關注 3.1K 粉絲

Wësche@WescheNex1q

2026.10.07 01:00

Le Chonk vs full GLM-5.3 on the same voxel pagoda prompt. -50K tokens vs 126K. 19.5 min vs 72. Both ran first try, untouched. Thinking on, temp 0.7, no output cap. GLM local on 4 Sparks, Mistral via @OpenRouterAI I think it did a pretty impressive job