Read news on ASR with our app.
Read more in the app
Cohere's open-weight ASR model hits 5.4% word error rate — low enough to replace speech APIs in production pipelines
Show HN: Sparrow-1 – Audio-native model for human-level turn-taking without ASR
