Comment by KerrAvon
1 day ago
If I'm doing the math right, that's like 17 t/s? I haven't played with Qwen 3.8 yet, but that seems really slow for a 27B on an M5 Max with sufficient RAM to hold it in memory.
1 day ago
If I'm doing the math right, that's like 17 t/s? I haven't played with Qwen 3.8 yet, but that seems really slow for a 27B on an M5 Max with sufficient RAM to hold it in memory.
No comments yet
Contribute on Hacker News ↗