>>531022310
Nowadays I assume it's code for "oh shit something went wrong" or "we need to make a change fast". L2 34B and WizardLM 2 both got put out of their misery by the whole "we need to do more red teaming" thing
I think it's more that the chink model got released at an unfortunate time for them and shit on their hype campaign of we're releasing the heckin best open model ever. I feel like Kimi probably comes close on their benchmarks which looks extra bad since it isn't even a reasoning model
