@moiiikaaa skilled engineers think and can still work if weekly tokens are used. slop engineers rely on super smart models, and considers used up tokens as a blocker.
@denchius_dev@rezoundous let alone defend it. People don't need super smart models. There are 2 other models that is on par with gpt and claude. Unless they don't want to think anymore and that is a problem, or they are solving cancer.
@bridgemindai tested this on my erp with million loc. muse spark 1.3 is stuck in a loop for 40minutes I stopped it. Switched back to glm 5.3 benchmarks doesn't mean anything now
@HeyAliux Let them stay with their god tier models. We don't want people flocking over other good models. Besides they think a model is all the answer to their problems. Not giving a hint if you create a god tier harness even qwen 3.6 35b can do what fable can do