Speech Arena
Elevenlabs V3 1,617 tops its leaderboard GPT-4o Mini TTS 1,566 leads the models we track
Not in model cards Models generate speech. Listeners compare outputs, and pairwise votes determine Elo ratings.
The Bradley-Terry view estimates the same vote pool and remains an auxiliary metric.
- Made by
- Speech Arena
- Kind
- Audio understanding and voice
- Level
- Everyday
- Scored
- Ranked by votes
- Answers
- Open-ended
- Tasks
- All public
- Last model added
- March 2025
- Scores refreshed
- November 12th, 2026
16
Entries on its own leaderboard, 1 model we track
Elevenlabs V3 tops the leaderboard with 1,617
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 TTS, at 1,566.
- 1,617 Elevenlabs V3
- 1,578 Cartesia Sonic Turbo
- 1,566 GPT-4o Mini TTS
- 1,532 Elevenlabs Flash 2.5
- 1,527 Cartesia Sonic 3
- 1,517 Murf AI Gen2
- 1,515 Elevenlabs eleven_multilingual_v2
- 1,504 Elevenlabs Turbo 2.5
- 1,493 InWorld Inworld TTS 1.5 Max
- 1,486 Open AI TTS-1
- 1,479 Open AI TTS-1-HD
- 1,453 Murf AI Falcon
- 1,451 Deepgram Aura 2
- 1,423 Sarvam Bulbul
- 1,414 InWorld Inworld TTS 1.5 Mini
- 1,364 Amazon Polly Amazon Polly Neural
Bars start at 1,300.
- 1 Elevenlabs V3 1,617
- 2 Cartesia Sonic Turbo 1,578
- 3 GPT-4o Mini TTS 1,566
- 4 Elevenlabs Flash 2.5 1,532
- 5 Cartesia Sonic 3 1,527
- 6 Murf AI Gen2 1,517
- 7 Elevenlabs eleven_multilingual_v2 1,515
- 8 Elevenlabs Turbo 2.5 1,504
- 9 InWorld Inworld TTS 1.5 Max 1,493
- 10 Open AI TTS-1 1,486
- 11 Open AI TTS-1-HD 1,479
- 12 Murf AI Falcon 1,453
- 13 Deepgram Aura 2 1,451
- 14 Sarvam Bulbul 1,423
- 15 InWorld Inworld TTS 1.5 Mini 1,414
- 16 Amazon Polly Amazon Polly Neural 1,364
Source: Official Speech Arena leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Elevenlabs V3 | 1,617 | Sep 2026 | Official leaderboard | |
| 2 | Cartesia Sonic Turbo | 1,578 | Oct 2026 | Official leaderboard | |
| 3 | GPT-4o Mini TTS | 1,566 | Mar 2025 | Official leaderboard | |
| 4 | Elevenlabs Flash 2.5 | 1,532 | Sep 2026 | Official leaderboard | |
| 5 | Cartesia Sonic 3 | 1,527 | Sep 2026 | Official leaderboard | |
| 6 | Murf AI Gen2 | 1,517 | Oct 2026 | Official leaderboard | |
| 7 | Elevenlabs eleven_multilingual_v2 | 1,515 | Feb 2026 | Official leaderboard | |
| 8 | Elevenlabs Turbo 2.5 | 1,504 | Feb 2026 | Official leaderboard | |
| 9 | InWorld Inworld TTS 1.5 Max | 1,493 | Oct 2026 | Official leaderboard | |
| 10 | Open AI TTS-1 | 1,486 | Sep 2026 | Official leaderboard | |
| 11 | Open AI TTS-1-HD | 1,479 | Oct 2026 | Official leaderboard | |
| 12 | Murf AI Falcon | 1,453 | Oct 2026 | Official leaderboard | |
| 13 | Deepgram Aura 2 | 1,451 | Sep 2026 | Official leaderboard | |
| 14 | Sarvam Bulbul | 1,423 | Oct 2026 | Official leaderboard | |
| 15 | InWorld Inworld TTS 1.5 Mini | 1,414 | Oct 2026 | Official leaderboard | |
| 16 | Amazon Polly Amazon Polly Neural | 1,364 | Oct 2026 | 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
Source: Official Speech Arena 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%