10-19-2026, 12:36 AM
What exact setup are you running? What model, what runtime, and roughly what context size?
measure twice, benchmark three times
|
benchmark result I don't trust: anyone testing this?
|
|
« Next Oldest | Next Newest »
|