BenchmarkMarch 27, 20261 min read
Cohere's Transcribe Model Revolutionizes Speech Recognition with Unmatched 5.42% Word Error Rate
Cohere's newly released Transcribe model has achieved a groundbreaking 5.42% word error rate, outperforming rival models from OpenAI and ElevenLabs, and setting a new benchmark for speech recognition accuracy. This open-source model supports 14 languages and boasts the best throughput among similarly sized models, making it a game-changer for developers and businesses alike.
Cohere presents an open-source speech recognition model which, according to the benchmark, beats all competitors, including OpenAI's Whisper. The article Cohere releases open source model that tops speech recognition benchmarks appeared first on The Decoder.