whisper alternatives lately on M3 Max

[cudascope]

Good point. That matches what I saw on B200. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[visionforge]

Need a little more detail here. Do you see the same thing with a smaller batch or shorter prompt?

[quiethouse]

Thanks, that helps. I ended up in about the same place using Ollama. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[vectorfield]

Thanks, that helps. I ended up in about the same place using llama.cpp. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.