Cekura Speech to Text
ElevenLabs Scribe v3 Live 1.89% tops its leaderboard gpt-4o-mini-transcribe 3.73% leads the models we track
Not in model cards Overall word error rate across the published audio clips.
Separate task board. No combined Cekura score exists. Costs use their explicitly named units.
- Made by
- Cekura
- Released
- July 2026
- Kind
- Audio understanding and voice
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Open-ended
- Tasks
- Public, plus a private test set
- Last model added
- March 2025
- Scores refreshed
- November 12th, 2026
21
Entries on its own leaderboard, 2 models we track
ElevenLabs Scribe v3 Live tops the leaderboard with 1.89%
Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is gpt-4o-mini-transcribe, at 3.73%.
- 1 ElevenLabs Scribe v3 Live 1.89%
- 2 Reson8 Rumble-2 2.05%
- 3 Fish Audio Transcribe Stream 1 2.21%
- 4 NVIDIA Nemotron 3.5 Transcribe 2.28%
- 5 Reson8 Resonant-1 2.38%
- 6 Cartesia Ink 2 2.63%
- 7 GPT Realtime Whisper 2.63%
- 8 AssemblyAI Universal 3.6 Pro 2.71%
- 9 AssemblyAI Universal 3.5 Pro 2.72%
- 10 Google Chirp 3 2.76%
- 11 Speechmatics Linden 2.97%
- 12 Google Chirp 2 3.01%
- 13 Soniox STT-RT v5 3.02%
- 14 Smallest Pulse 3.59%
- 15 gpt-4o-mini-transcribe 3.73%
- 16 Deepgram Nova-3 3.78%
- 17 Deepgram Flux English 4.00%
- 18 Deepgram Flux Multilingual 4.81%
- 19 Inworld STT-1 5.44%
- 20 Gradium 5.78%
- 21 gpt-4o-transcribe 6.85%
Source: Official Cekura Speech to Text leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | ElevenLabs Scribe v3 Live | 1.89% | Sep 2026 | Official leaderboard | |
| 2 | Reson8 Rumble-2 | 2.05% | Sep 2026 | Official leaderboard | |
| 3 | Fish Audio Transcribe Stream 1 | 2.21% | Sep 2026 | Official leaderboard | |
| 4 | NVIDIA Nemotron 3.5 Transcribe | 2.28% | Sep 2026 | Official leaderboard | |
| 5 | Reson8 Resonant-1 | 2.38% | Sep 2026 | Official leaderboard | |
| 6 | Cartesia Ink 2 | 2.63% | Sep 2026 | Official leaderboard | |
| 7 | GPT Realtime Whisper | 2.63% | Sep 2026 | Official leaderboard | |
| 8 | AssemblyAI Universal 3.6 Pro | 2.71% | Sep 2026 | Official leaderboard | |
| 9 | AssemblyAI Universal 3.5 Pro | 2.72% | Sep 2026 | Official leaderboard | |
| 10 | Google Chirp 3 | 2.76% | Sep 2026 | Official leaderboard | |
| 11 | Speechmatics Linden | 2.97% | Sep 2026 | Official leaderboard | |
| 12 | Google Chirp 2 | 3.01% | Sep 2026 | Official leaderboard | |
| 13 | Soniox STT-RT v5 | 3.02% | Sep 2026 | Official leaderboard | |
| 14 | Smallest Pulse | 3.59% | Sep 2026 | Official leaderboard | |
| 15 | gpt-4o-mini-transcribe | 3.73% | Mar 2025 | Official leaderboard | |
| 16 | Deepgram Nova-3 | 3.78% | Sep 2026 | Official leaderboard | |
| 17 | Deepgram Flux English | 4.00% | Sep 2026 | Official leaderboard | |
| 18 | Deepgram Flux Multilingual | 4.81% | Sep 2026 | Official leaderboard | |
| 19 | Inworld STT-1 | 5.44% | Sep 2026 | Official leaderboard | |
| 20 | Gradium | 5.78% | Sep 2026 | Official leaderboard | |
| 21 | gpt-4o-transcribe | 6.85% | Mar 2025 | Official leaderboard |
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
On the benchmark’s own leaderboard
Best so far
Source: Official Cekura Speech to Text leaderboard
No lab model card reports it yet
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.
More in Audio and video
- Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy.
Qwen 3.8 Max 88.7% - LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization.
Gemini 4 Argon 91.7% - MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy. Qwen 3.8 Max 82.4%
- MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy. Qwen 3.8 Max 90.8%
- FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better.
Phi-4-multimodal 4.00% - VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules. Qwen3.5-Omni-Plus 93.1%