Read the docs

Sophea K1 on the
Open ASR Leaderboard.

An independent ranking of open and proprietary speech-to-text models by word error rate. Sophea K1 ranks fourth in German and seventh in English.

Transcribe with K1See the leaderboard
WHERE K1 STANDS

Among the leading speech models.

Hugging Face's Open ASR Leaderboard, results of 26 October 2026. Lower word error rate is better.

German

FLEURS · Common Voice

#4Fourth place

2.54%average WER

  1. 1
  2. 2
  3. 3
  4. K1
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10

0.62 points behind the leader, ahead of 6 of the top ten.

Open ASR Leaderboard

English

11 English test sets

#7Seventh place

5.23%average WER

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. K1
  8. 8
  9. 9
  10. 10

0.68 points behind the leader, ahead of 3 of the top ten.

Open ASR Leaderboard
THE TOP TEN

The boards, as published.

Each language's ten best models by average word error rate, with K1 marked. Shorter bars are better.

German average word error rate

RankModelWER %
1microsoft/azure-speech-07-20261.92
2elevenlabs/scribe_v22.24
3assemblyai/universal-3-5-pro2.43
4sophea/asr-k1 (preview)Sophea2.54
5assemblyai/universal-3-pro2.59
6sprag/symphony2.72
7reson8/resonant-12.78
8reson8/resonant-1-flash2.78
9speechmatics/enhanced2.78
10openai/gpt-transcribe2.84

English average word error rate

RankModelWER %
1OtterAI/speech-v3.14.55
2zoom/scribe_v2_pro4.70
3p99lab/audrey-0.74.98
4microsoft/azure-speech-07-20265.02
5modulate/multilingual5.06
6reson8/resonant-15.21
7sophea/asr-k1 (preview)Sophea5.23
8elevenlabs/scribe_v25.24
9assemblyai/universal-3-5-pro5.39
10Qwen/Qwen3-ASR-1.7B-hf5.40

Source: Open ASR Leaderboard by Hugging Face (opens in a new tab), results last updated 26 October 2026, read 28 October 2026. K1 is listed there as sophea/asr-k1 (preview). Ranks change as models are added. K1 has no results yet for the leaderboard's other languages.