Skip to content
Model Wayfinder
← Back to comparison
Cohere Labs

cohere-transcribe-03-2026

Open weightsApache 2.0 · commercial use allowed

Benchmarks

Benchmark average
75/100
based on 1 benchmarks
BenchmarkScorePositionSource
Open ASR (English) · WER speech recognition (word error rate)6.72 %75/ 69Hugging Face Open ASR Leaderboard · Nov 17, 2026

Position: 100 = best, 0 = last among all models listed here that were tested on this benchmark.

Strengths & weaknesses

Automatic summary from benchmark positions – not an opinion.

Pros
  • +Strong in speech recognition (word error rate) (position 75/100)
Cons

    Our quality rating

    Transcribing audio3/3

    Tiers derived from: Open ASR Leaderboard 2026-11-17

    Source, checked Nov 17, 2026