What's "512k"?
#1
by
tokyohandsome
- opened
What's the difference from your previous conversion: nightmedia/Qwen3-30B-A3B-Thinking-2507-dwq4-mlx ?
The original model context length is 256k, I roped it 2x to 512k
It seems to exhibit different planning when it “sees” long context, not that it needs to go that far
I created it to be able to summarize a long context that already hit that limit.
Ah, I see, thanks.