ElevenLabsAudio modelActive

Eleven Flash v2.5

PopularityRank 569 of 875

22out of 100same as last month

Summary

Eleven Flash v2.5 is designed for low-latency text-to-speech output. ElevenLabs lists about 75 ms latency, 32 languages, and up to 40,000 characters per request.

Flash v2.5 normalizes phone numbers, dates, and currencies less reliably by default than slower models. The documentation lists normalization as an Enterprise option. Its low latency does not remove that documented boundary for numeric text.

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 234 published benchmark results

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

Coval pooled corpus-level synthesized-speech word error rate (TTS WER) (Show measurement, test conditions, and source)

Source value: 1.8509058333148243
Score: 1.8509058333148243
Metric: Pooled corpus-level synthesized-speech word error rate, Coval pooled_value
Unit: %
Benchmark version: Coval rolling 30-day window, re-measured daily; route last measured 2026-10-16
Category: audio
Direction: Lower is better
Sample: 2090
Source type: official-leaderboard
Evaluator: Coval
Status: superseded
Retrieved at: 2026-10-16
Methodology: Official benchmark leaderboard. Coval endpoint key elevenlabs/eleven_flash_v2_5, exact ElevenLabs API model ID eleven_flash_v2_5, hosted by ElevenLabs. This is a model-scoped synthesized-speech WER observation; the public route does not identify the tested voice. It is not STT WER. Rolling 30-day window, re-measured daily.
Context: This is Coval's pooled corpus-level synthesized-speech WER for the exact Eleven Flash v2.5 endpoint, not the separate per-clip mean_value or a rounded leaderboard headline. Pooled WER and mean per-clip WER are different aggregations and should not be compared as the same score. The WER row's sample_count is used, not another metric's count or a roster total. It measures output intelligibility, not STT accuracy. The tested voice is not identified by the public route, so this is not a voice-specific result. Results are from Coval's hosted ElevenLabs endpoint, not local or native performance.

Source Coval

MiniMaxAI/TTS-Multilingual-Test-Set WER, average (Show measurement, test conditions, and source)

Source value: 1.52
Score: 1.52
Metric: Word error rate (WER); lower is better
Unit: %
Benchmark version: Gradium evaluation on MiniMaxAI/TTS-Multilingual-Test-Set, evaluated 2026-04; report published 2026-04-29; Qwen3-ASR
Direction: Lower is better
Source type: vendor-reported
Evaluator: Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. MiniMaxAI/TTS-Multilingual-Test-Set (24 available languages; five evaluated: English, French, Spanish, Portuguese, German); reference STT was Qwen3-ASR. The English normalizer was Whisper English, French used kyutai/tts_longeval, and Spanish, Portuguese, and German used Whisper basic. The report identifies ElevenLabs Flash v2.5 by product name; utterance count, voice IDs, API revision, and sampling controls are not published.
Context: Gradium is a competing TTS vendor and labels this table self-reported. Only five of the 24 dataset languages were evaluated. The table does not publish metric-level N, confidence intervals, immutable serving revision, voice IDs, or API controls. WER depends on reference ASR and language normalizers, so this row is not directly comparable with other WER series.

Source Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy

MiniMaxAI/TTS-Multilingual-Test-Set WER, English (Show measurement, test conditions, and source)

Source value: 0.36
Score: 0.36
Metric: Word error rate (WER); lower is better
Unit: %
Benchmark version: Gradium evaluation on MiniMaxAI/TTS-Multilingual-Test-Set, evaluated 2026-04; report published 2026-04-29; Qwen3-ASR
Direction: Lower is better
Source type: vendor-reported
Evaluator: Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. MiniMaxAI/TTS-Multilingual-Test-Set (24 available languages; five evaluated: English, French, Spanish, Portuguese, German); reference STT was Qwen3-ASR. The English normalizer was Whisper English, French used kyutai/tts_longeval, and Spanish, Portuguese, and German used Whisper basic. The report identifies ElevenLabs Flash v2.5 by product name; utterance count, voice IDs, API revision, and sampling controls are not published.
Context: Gradium is a competing TTS vendor and labels this table self-reported. Only five of the 24 dataset languages were evaluated. The table does not publish metric-level N, confidence intervals, immutable serving revision, voice IDs, or API controls. WER depends on reference ASR and language normalizers, so this row is not directly comparable with other WER series.

Source Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy

MiniMaxAI/TTS-Multilingual-Test-Set WER, French (Show measurement, test conditions, and source)

Source value: 2.45
Score: 2.45
Metric: Word error rate (WER); lower is better
Unit: %
Benchmark version: Gradium evaluation on MiniMaxAI/TTS-Multilingual-Test-Set, evaluated 2026-04; report published 2026-04-29; Qwen3-ASR
Direction: Lower is better
Source type: vendor-reported
Evaluator: Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. MiniMaxAI/TTS-Multilingual-Test-Set (24 available languages; five evaluated: English, French, Spanish, Portuguese, German); reference STT was Qwen3-ASR. The English normalizer was Whisper English, French used kyutai/tts_longeval, and Spanish, Portuguese, and German used Whisper basic. The report identifies ElevenLabs Flash v2.5 by product name; utterance count, voice IDs, API revision, and sampling controls are not published.
Context: Gradium is a competing TTS vendor and labels this table self-reported. Only five of the 24 dataset languages were evaluated. The table does not publish metric-level N, confidence intervals, immutable serving revision, voice IDs, or API controls. WER depends on reference ASR and language normalizers, so this row is not directly comparable with other WER series.

Source Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy

MiniMaxAI/TTS-Multilingual-Test-Set WER, Spanish (Show measurement, test conditions, and source)

Source value: 0.99
Score: 0.99
Metric: Word error rate (WER); lower is better
Unit: %
Benchmark version: Gradium evaluation on MiniMaxAI/TTS-Multilingual-Test-Set, evaluated 2026-04; report published 2026-04-29; Qwen3-ASR
Direction: Lower is better
Source type: vendor-reported
Evaluator: Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. MiniMaxAI/TTS-Multilingual-Test-Set (24 available languages; five evaluated: English, French, Spanish, Portuguese, German); reference STT was Qwen3-ASR. The English normalizer was Whisper English, French used kyutai/tts_longeval, and Spanish, Portuguese, and German used Whisper basic. The report identifies ElevenLabs Flash v2.5 by product name; utterance count, voice IDs, API revision, and sampling controls are not published.
Context: Gradium is a competing TTS vendor and labels this table self-reported. Only five of the 24 dataset languages were evaluated. The table does not publish metric-level N, confidence intervals, immutable serving revision, voice IDs, or API controls. WER depends on reference ASR and language normalizers, so this row is not directly comparable with other WER series.

Source Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy

MiniMaxAI/TTS-Multilingual-Test-Set WER, Portuguese (Show measurement, test conditions, and source)

Source value: 3.18
Score: 3.18
Metric: Word error rate (WER); lower is better
Unit: %
Benchmark version: Gradium evaluation on MiniMaxAI/TTS-Multilingual-Test-Set, evaluated 2026-04; report published 2026-04-29; Qwen3-ASR
Direction: Lower is better
Source type: vendor-reported
Evaluator: Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. MiniMaxAI/TTS-Multilingual-Test-Set (24 available languages; five evaluated: English, French, Spanish, Portuguese, German); reference STT was Qwen3-ASR. The English normalizer was Whisper English, French used kyutai/tts_longeval, and Spanish, Portuguese, and German used Whisper basic. The report identifies ElevenLabs Flash v2.5 by product name; utterance count, voice IDs, API revision, and sampling controls are not published.
Context: Gradium is a competing TTS vendor and labels this table self-reported. Only five of the 24 dataset languages were evaluated. The table does not publish metric-level N, confidence intervals, immutable serving revision, voice IDs, or API controls. WER depends on reference ASR and language normalizers, so this row is not directly comparable with other WER series.

Source Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy

MiniMaxAI/TTS-Multilingual-Test-Set WER, German (Show measurement, test conditions, and source)

Source value: 0.61
Score: 0.61
Metric: Word error rate (WER); lower is better
Unit: %
Benchmark version: Gradium evaluation on MiniMaxAI/TTS-Multilingual-Test-Set, evaluated 2026-04; report published 2026-04-29; Qwen3-ASR
Direction: Lower is better
Source type: vendor-reported
Evaluator: Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy
Status: active
Retrieved at: 2026-10-17
Methodology: Vendor-reported result. MiniMaxAI/TTS-Multilingual-Test-Set (24 available languages; five evaluated: English, French, Spanish, Portuguese, German); reference STT was Qwen3-ASR. The English normalizer was Whisper English, French used kyutai/tts_longeval, and Spanish, Portuguese, and German used Whisper basic. The report identifies ElevenLabs Flash v2.5 by product name; utterance count, voice IDs, API revision, and sampling controls are not published.
Context: Gradium is a competing TTS vendor and labels this table self-reported. Only five of the 24 dataset languages were evaluated. The table does not publish metric-level N, confidence intervals, immutable serving revision, voice IDs, or API controls. WER depends on reference ASR and language normalizers, so this row is not directly comparable with other WER series.

Source Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy

Async TTS pronunciation benchmark, ElevenLabs Flash v2.5 sentence-level normalization accuracy (Show measurement, test conditions, and source)

Source value: 40.3
Score: 40.3
Metric: Strict sentence-level text-normalization accuracy; every normalization unit in a sentence must be pronounced correctly; higher is better
Unit: %
Benchmark version: Async TTS pronunciation benchmark, full article captured 2026-10-17
Category: audio
Direction: Higher is better
Sample: 1022
Source type: independent-evaluator
Evaluator: Async TTS Pronunciation Benchmark
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. The benchmark names ElevenLabs Flash v2.5, mapped to API route eleven_flash_v2_5 in the current official model overview. Audio is generated through the real-time WebSocket streaming interface from raw text without cleanup or LLM rewriting; Gemini 3.1 Pro applies category-specific normalization rules. This row is the published score for the named evaluation dimension.
Context: The independent page reports no confidence interval, exact voice selection or immutable provider serving revision. Expert linguists reviewed about 300 sentence/model pairs overall; the page does not publish a model-specific validation estimate.

Source Async TTS Pronunciation Benchmark

Async TTS pronunciation benchmark, ElevenLabs Flash v2.5 unit-level normalization accuracy (Show measurement, test conditions, and source)

Source value: 56.4
Score: 56.4
Metric: Accuracy over individual non-standard text-normalization units; higher is better
Unit: %
Benchmark version: Async TTS pronunciation benchmark, full article captured 2026-10-17
Category: audio
Direction: Higher is better
Sample: 2243
Source type: independent-evaluator
Evaluator: Async TTS Pronunciation Benchmark
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. The benchmark names ElevenLabs Flash v2.5, mapped to API route eleven_flash_v2_5 in the current official model overview. Audio is generated through the real-time WebSocket streaming interface from raw text without cleanup or LLM rewriting; Gemini 3.1 Pro applies category-specific normalization rules. This row is the published score for the named evaluation dimension.
Context: The independent page reports no confidence interval, exact voice selection or immutable provider serving revision. Expert linguists reviewed about 300 sentence/model pairs overall; the page does not publish a model-specific validation estimate.

Source Async TTS Pronunciation Benchmark

Cekura TTS hard-part accuracy: numbers and shorthand (Show measurement, test conditions, and source)

Source value: 55
Score: 55
Metric: Share of hard-part readings accepted for numbers and shorthand
Unit: %
Benchmark version: Cekura TTS Model Benchmarks; current page snapshot 2026-10-19; run date and serving revision not stated
Category: audio
Direction: Higher is better
Source type: official-leaderboard
Evaluator: Cekura Bench TTS Model Benchmarks
Status: active
Retrieved at: 2026-10-19
Methodology: Official benchmark leaderboard. Evaluator-run TTS benchmark. Cekura says each model read the same 104 agent sentences three times from one US East cloud region, using one stock voice and its own streaming protocol. Non-streaming timing runs from sending the full text to first audible sound; streaming timing runs from the last word sent to first audible sound. Reading accuracy is estimated from Deepgram Nova-3 and OpenAI Whisper transcripts, counting accepted readings without human review. Naturalness uses UTMOS, an open model predicting a listening-test score. The captured rows do not state per-cell N, CI, exact run date, serving/API revision, or exact voice ID unless shown separately.
Context: Cekura discloses that it runs the benchmark itself and sells testing and monitoring tools in the same market. The results are evaluator-published and include voice, runtime, protocol, ASR, and UTMOS effects; they are not a human-only or isolated-model evaluation.

Source Cekura Bench TTS Model Benchmarks

Cekura TTS hard-part accuracy: codes and addresses (Show measurement, test conditions, and source)

Source value: 55
Score: 55
Metric: Share of hard-part readings accepted for codes and addresses
Unit: %
Benchmark version: Cekura TTS Model Benchmarks; current page snapshot 2026-10-19; run date and serving revision not stated
Category: audio
Direction: Higher is better
Source type: official-leaderboard
Evaluator: Cekura Bench TTS Model Benchmarks
Status: active
Retrieved at: 2026-10-19
Methodology: Official benchmark leaderboard. Evaluator-run TTS benchmark. Cekura says each model read the same 104 agent sentences three times from one US East cloud region, using one stock voice and its own streaming protocol. Non-streaming timing runs from sending the full text to first audible sound; streaming timing runs from the last word sent to first audible sound. Reading accuracy is estimated from Deepgram Nova-3 and OpenAI Whisper transcripts, counting accepted readings without human review. Naturalness uses UTMOS, an open model predicting a listening-test score. The captured rows do not state per-cell N, CI, exact run date, serving/API revision, or exact voice ID unless shown separately.
Context: Cekura discloses that it runs the benchmark itself and sells testing and monitoring tools in the same market. The results are evaluator-published and include voice, runtime, protocol, ASR, and UTMOS effects; they are not a human-only or isolated-model evaluation.

Source Cekura Bench TTS Model Benchmarks

Cekura TTS hard-part accuracy: read from context (Show measurement, test conditions, and source)

Source value: 75
Score: 75
Metric: Share of hard-part readings accepted for context-dependent readings
Unit: %
Benchmark version: Cekura TTS Model Benchmarks; current page snapshot 2026-10-19; run date and serving revision not stated
Category: audio
Direction: Higher is better
Source type: official-leaderboard
Evaluator: Cekura Bench TTS Model Benchmarks
Status: active
Retrieved at: 2026-10-19
Methodology: Official benchmark leaderboard. Evaluator-run TTS benchmark. Cekura says each model read the same 104 agent sentences three times from one US East cloud region, using one stock voice and its own streaming protocol. Non-streaming timing runs from sending the full text to first audible sound; streaming timing runs from the last word sent to first audible sound. Reading accuracy is estimated from Deepgram Nova-3 and OpenAI Whisper transcripts, counting accepted readings without human review. Naturalness uses UTMOS, an open model predicting a listening-test score. The captured rows do not state per-cell N, CI, exact run date, serving/API revision, or exact voice ID unless shown separately.
Context: Cekura discloses that it runs the benchmark itself and sells testing and monitoring tools in the same market. The results are evaluator-published and include voice, runtime, protocol, ASR, and UTMOS effects; they are not a human-only or isolated-model evaluation.

Source Cekura Bench TTS Model Benchmarks

Page 1 of 20

Popularity compared

How often the model is searched for, used and mentioned, compared with every model of its category.

Popularity index

0 to 100, higher is more popular

20 of 83 models

pyannote Community-173

Nemotron-3-Diarization57

Gemini 3.5 Transcribe52

Confucius4-R2T252

Granite Speech 5.0 470M TurboCTC52

Kokoro-82M51

Nova-351

Fish Audio S2 Pro51

NVIDIA Nemotron 3.5 Transcribe50

Granite Speech 5.0 470M TurboCTC NC50

Orukeet49

Lyria 3.547

Gemini 3.5 Live Translate46

Phonon-244

Aura-244

Gemini 3.1 Flash TTS44

MiniMax-Music341

Audio8-ASR-Infinite41

GPT-Realtime 2.141

Eleven Flash v2.522

pyannote Community-1

Nemotron-3-Diarization

Gemini 3.5 Transcribe

Confucius4-R2T2

Granite Speech 5.0 470M TurboCTC

Kokoro-82M

Nova-3

Fish Audio S2 Pro

NVIDIA Nemotron 3.5 Transcribe

Granite Speech 5.0 470M TurboCTC NC

Orukeet

Lyria 3.5

Gemini 3.5 Live Translate

Eleven Flash v2.5

As of November 2026 · Source: gradually.ai popularity index

Over time

Weekly values from October 18, 2026 to November 5, 2026

Eleven Flash v2.5Index-Echo-S2TT-9BLyria RealtimeIndex-Echo-S2TT-2BLyria 3 Clip

All specifications

Architecture

Model class: Text-to-speech; Source: Model class

Input and output

Input: Text; Source: Input
Output: Audio; Source: Output
Languages: 32 languages; Source: Languages

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking: Subject to limitations. ElevenLabs describes an inaudible watermark in its own generated audio and identifies SynthID, including for speech generation. Older recordings made before its introduction and third-party model outputs are not automatically covered.; Source 1: AI watermarking Source 2: AI watermarking Source 3: AI watermarking Source 4: AI watermarking
Watermarking technique: SynthID; Source 1: Watermarking technique Source 2: Watermarking technique Source 3: Watermarking technique Source 4: Watermarking technique
Labelling reviewed: Nov 5, 2026; Source 1: Labelling reviewed Source 2: Labelling reviewed Source 3: Labelling reviewed Source 4: Labelling reviewed

Licence and access

Access: API; Source: Access

API

API model ID: eleven_flash_v2_5; Source: API model ID

Further details

Task: Low-latency speech generation; Source: Task
Supported details: According to ElevenLabs, about 75 ms latency and 32 languages; Source: Supported details
Text input: Up to 40,000 characters; Source: Text input

Sources and data date

Show all 45 sources

Benchmark sources at a glance

Which independent evaluators tested this model. This is coverage, not a quality ranking.

No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.

Without vendor claims and historical or disputed results; those stay labelled in the single values.

Type	Evidence and data date
Research status	Research date October 15, 2026. 36 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not published	ttsbench-elevenlabs-serving-revision Original wording The source config specifies eleven_flash_v2_5 and its voice and API controls but does not bind an immutable ElevenLabs checkpoint or serving revision.
Unresolved	exact-serving-snapshot-and-unpublished-method-details Note The captured external source reports a benchmark-specific model or hosted route but does not establish an immutable currently served checkpoint. Unpublished row-level denominators and confidence intervals remain unknown; sample counts were entered only for cells that state them.
Additional source	ElevenLabs model documentation (retrieved September 24, 2026; October 15, 2026; October 17, 2026) · ElevenLabs Models documentation · Editorial description reviewed October 2, 2026
Additional source	Artificial Analysis (retrieved October 15, 2026; October 17, 2026; November 3, 2026)
Additional source	SpeechArena (retrieved October 15, 2026)
Additional source	Coval (retrieved October 16, 2026; October 17, 2026; October 20, 2026; October 21, 2026; October 28, 2026; October 29, 2026; October 30, 2026; October 31, 2026; November 1, 2026; November 2, 2026; November 5, 2026; November 6, 2026)
Additional source	Rime Labs TTSBench (retrieved October 17, 2026)1 2
Additional source	Gradium, TTS WER Benchmark 2026: Pronunciation Accuracy (retrieved October 17, 2026)
Additional source	Async TTS Pronunciation Benchmark (retrieved October 17, 2026)
Additional source	Breeze Blue TTS Latency Benchmark (retrieved October 18, 2026)
Additional source	SocialPro (retrieved October 18, 2026)
Additional source	Cekura Bench TTS Model Benchmarks (retrieved October 19, 2026)
Additional source	Artificial Analysis (retrieved October 28, 2026; October 29, 2026; October 30, 2026; November 3, 2026)1 2
Additional source	Speko (retrieved October 28, 2026)
Additional source	Model metadata (retrieved October 16, 2026)1 2
Additional source	Model metadata (retrieved October 17, 2026)
Additional source	Model metadata (retrieved October 18, 2026)
Additional source	Model metadata (retrieved October 28, 2026)1 2 3 4 5 6 7 8 9 10 11
Additional source	Model metadata (retrieved November 5, 2026)1 2 3
Additional source	Model metadata (retrieved November 6, 2026)1 2 3
Additional source	ElevenLabs: Audio Detector (retrieved November 5, 2026)
Additional source	ElevenLabs: Audio provenance (retrieved November 5, 2026)
Additional source	ElevenLabs: Voice generation API (retrieved November 5, 2026)
Additional source	ElevenLabs: Audio Detector and classifier (retrieved November 5, 2026)
Additional source	OpenRouter: Popularity index (derived), CC BY 4.0
Additional source	Vercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional source	Ollama: cumulative downloads for the popularity index
Additional source	Replicate: cumulative runs for the popularity index
Additional source	Google Trending Now: limited feed selection for Germany and the US
Additional source	Creative Commons Attribution 4.0