GoogleAudio modelActive
Gemini 3.5 Transcribe Live
PopularityRank 358 of 875
35out of 100same as last month
Summary
Gemini 3.5 Transcribe Live brings the same transcription family to streaming. It accepts an audio stream, produces text with low latency, recognizes more than 85 languages, and handles code-switching during a session. Smart Dictation and custom vocabulary of up to 1,000 terms extend the live transcript with editorial hints.
A live session lasts up to 10 minutes, and live mode does not support speaker diarization or word-level timestamps. Gemini 3.5 Transcribe is the file endpoint for those annotations, including speaker assignment and word-level alignment.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 183 published benchmark results
Page 1 of 16
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.
Artificial Analysis AA-WER Streaming (final transcription) (Show measurement, test conditions, and source)
- Source value
- 3.9983
- Score
- 3.9983
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- AA-WER v2.2 streaming
- Category
- audio
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Independent evaluation. Direct API streaming evaluation; final-transcription WER across AA-AgentTalk (50%), VoxPopuli (25%), and Earnings22 (25%), about eight hours total. WER is duration-weighted within dataset; Earnings22 chunking is reassembled at parent-call level. AA-WER v2.2 streaming protocol.
Coval rolling-30-day pooled word error rate (Gemini 3.5 Transcribe Live endpoint) (Show measurement, test conditions, and source)
- Source value
- 2.9814359439806655
- Score
- 2.9814359439806655
- Metric
- Pooled word error rate across Coval's fixed public speech-recognition audio conditions
- Unit
- %
- Benchmark version
- Coval rolling 30-day snapshot, last measured 2026-10-16, source response fetched 2026-10-16T15:36:04.093476Z
- Category
- audio
- Direction
- Lower is better
- Sample
- 26595
- Source type
- independent-evaluator
- Evaluator
- Coval
- Status
- superseded
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact hosted endpoint Gemini 3.5 Transcribe Live. Coval reports daily fixed-public-audio endpoint tests over a rolling 30-day window, last measured 2026-10-16. Stored score is pooled_value=2.9814359439806655 from the WER row, not mean_value; Coval's visible 3.0% is rounded to one decimal. The reported WER sample count comes from the same row.
- Context
- The page reports a metric-specific scored sample count, not total attempted requests or failures, and reports no confidence interval. This is Coval's rolling hosted-endpoint observation, not an Artificial Analysis AA-WER result.
Google DeepMind FLEURS top-locales streaming word error rate (Show measurement, test conditions, and source)
- Source value
- 5.5
- Score
- 5.5
- Metric
- Word error rate across the FLEURS top-locales streaming index
- Unit
- %
- Benchmark version
- Google DeepMind FLEURS top-locales streaming index, accessed 2026-10-16
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Google DeepMind
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Google DeepMind reports 5.50% WER for Gemini 3.5 Transcribe Live in streaming mode on its FLEURS top-locales index. The official chart footnote lists ar-EG, bn-IN, cmn-Hans-CN, de-DE, en-US, es-ES, es-US, fr-FR, hi-IN, id-ID, it-IT, ja-JP, ko-KR, mr-IN, nl-NL, pl-PL, pt-BR, ro-RO, ru-RU, ta-IN, te-IN, th-TH, tr-TR, uk-UA, and vi-VN. The source does not specify the aggregation formula or evaluated sample count.
- Context
- Google publishes no evaluated N, confidence interval, per-locale scores, or aggregation formula for this index. This is a vendor-reported aggregate and is not interchangeable with Artificial Analysis or other WER results.
Pipecat STT Benchmark semantic WER mean (Show measurement, test conditions, and source)
- Source value
- 1.93
- Score
- 1.93
- Metric
- Arithmetic mean semantic word error rate across successfully transcribed samples
- Unit
- %
- Benchmark version
- Pipecat STT Benchmark at commit 94f567299a21f8d617039bc69f137a131f5dedd5, September 2026 semantic-WER judge update
- Category
- audio
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- Pipecat STT Benchmark
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. The pinned results table reports semantic WER over 1000 attempted samples from pipecat-ai/smart-turn-data-v3.1-train. WER Mean is the average across samples with a returned transcript; the table reports 99.9% transcript success, but does not give the exact number of WER-scored samples. Semantic scoring by Claude Sonnet 5.5 ignores punctuation, contractions, fillers, and equivalent phrasings. The source does not document a language distribution; its dataset preview shows English transcripts.
- Context
- No exact WER-scored N or model-specific confidence interval is published. The 1000 figure is attempted samples, not the WER denominator. The README says differences below about 0.25 percentage points are within the 95% bootstrap interval among the most accurate services, so do not rank close scores.
Pipecat STT Benchmark semantic pooled WER (Show measurement, test conditions, and source)
- Source value
- 1.97
- Score
- 1.97
- Metric
- Pooled semantic word error rate, total judged errors divided by total reference words
- Unit
- %
- Benchmark version
- Pipecat STT Benchmark at commit 94f567299a21f8d617039bc69f137a131f5dedd5, September 2026 semantic-WER judge update
- Category
- audio
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- Pipecat STT Benchmark
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. The pinned results table reports semantic WER over 1000 attempted samples from pipecat-ai/smart-turn-data-v3.1-train. Pooled WER is total judged semantic word errors divided by total reference words, among samples with a returned transcript; the table reports 99.9% transcript success but not the exact scored N. Semantic scoring by Claude Sonnet 5.5 ignores punctuation, contractions, fillers, and equivalent phrasings. The source does not document a language distribution; its dataset preview shows English transcripts.
- Context
- No exact WER-scored N or model-specific confidence interval is published. The 1000 figure is attempted samples, not the WER denominator. The README says differences below about 0.25 percentage points are within the 95% bootstrap interval among the most accurate services, so do not rank close scores.
Coval rolling-30-day pooled word error rate, 2026-10-17 snapshot (Show measurement, test conditions, and source)
- Source value
- 3
- Score
- 3
- Metric
- Pooled endpoint word error rate (WER)
- Unit
- %
- Benchmark version
- Coval daily fixed-audio endpoint benchmark, rolling 30-day window; route last measured 2026-10-17
- Category
- audio
- Direction
- Lower is better
- Sample
- 26580
- Source type
- independent-evaluator
- Evaluator
- Coval Voice AI Benchmarks
- Status
- superseded
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact hosted endpoint Gemini 3.5 Transcribe Live by Google. Coval remeasures fixed public speech audio daily and aggregates over a rolling 30-day window. The 2026-10-17 page reports 3.0% WER on 26580 samples, rounded to one decimal place.
- Context
- This is a new dated rolling-window snapshot, not a higher-precision replacement for the 2026-10-16 snapshot. The visible score is rounded to one decimal; no confidence interval is shown.
AssemblyAI normalized WER mean across three noise conditions (Show measurement, test conditions, and source)
- Source value
- 8.48
- Score
- 8.48
- Metric
- normalized word error rate, direct reported mean over no, low and heavy noise
- Unit
- %
- Benchmark version
- AssemblyAI benchmark snapshot; runs 2026-08-27 to 2026-09-22
- Category
- speech-to-text
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- AssemblyAI speech benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. AssemblyAI's benchmark page directly reports the exact row Gemini 3.5 Transcribe Live, with its mean of three conditions, no noise, low noise and heavy noise. The displayed row is 8.48%, with condition values 7.44%, 8.49% and 9.52%. Run window: 2026-08-27 to 2026-09-22; the page states one row per model, most recent run. The bundle records the published mean, not a recomputation.
- Context
- AssemblyAI is the evaluator and also a speech vendor. No per-model N, CI or immutable competitor serving revision is published.
AssemblyAI normalized medical entity error rate mean across three sets (Show measurement, test conditions, and source)
- Source value
- 8.57
- Score
- 8.57
- Metric
- normalized medical entity error rate, direct reported mean over three clinical sets
- Unit
- %
- Benchmark version
- AssemblyAI benchmark snapshot; runs 2026-08-18 to 2026-09-22
- Category
- medical speech-to-text
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- AssemblyAI speech benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. AssemblyAI's medical-entity table directly reports the exact Gemini 3.5 Transcribe Live row and the mean over primary care, clinical dictation and respiratory sets. The displayed mean is 8.57%, with 15.45%, 7.69% and 2.56% for those sets. Run window: 2026-08-18 to 2026-09-22; the page states one row per model, most recent run.
- Context
- The row is vendor-published comparative data. No per-model N, CI or immutable competitor serving revision is published.
Coval Gemini 3.5 Transcribe Live WER, run 20:00 UTC (Show measurement, test conditions, and source)
- Source value
- 3
- Score
- 3
- Metric
- Pooled fixed-public-audio word error rate (WER), page-rounded
- Unit
- %
- Benchmark version
- Coval daily public evaluator run; rolling 30-day window; last measured 2026-10-17 20:00 UTC
- Category
- audio
- Direction
- Lower is better
- Sample
- 26592
- Source type
- independent-evaluator
- Evaluator
- Coval Voice AI Benchmarks
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Coval public fixed-audio evaluation, rolling 30-day window and daily run. Exact page route/model label: Google Gemini 3.5 Transcribe Live endpoint; created by Google, hosted by Google Gemini API. Reported values and metric-specific denominators are retained; milliseconds are converted exactly to seconds. Last run stated on page: 2026-10-17 20:00 UTC.
- Context
- Coval publishes rolling 30-day results from daily repeated runs on fixed public audio or prompt sets. These rows are the later timestamped run visible in the captured page, not a copy of the earlier run in the input. Exact API/checkpoint revision, request parameters, and confidence intervals are not published; displayed WER is rounded to one decimal and is not a measure of naturalness.
Coval Gemini 3.5 Transcribe Live WER p95, run 20:00 UTC (Show measurement, test conditions, and source)
- Source value
- 16.7
- Score
- 16.7
- Metric
- 95th-percentile per-clip speech-recognition word error rate (WER)
- Unit
- %
- Benchmark version
- Coval daily public evaluator run; rolling 30-day window; last measured 2026-10-17 20:00 UTC
- Category
- speech-to-text
- Direction
- Lower is better
- Sample
- 26592
- Source type
- independent-evaluator
- Evaluator
- Coval Voice AI Benchmarks
- Status
- superseded
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Coval public fixed-audio evaluation, rolling 30-day window and daily run. Exact page route/model label: Google Gemini 3.5 Transcribe Live endpoint; created by Google, hosted by Google Gemini API. Reported values and metric-specific denominators are retained; milliseconds are converted exactly to seconds. Last run stated on page: 2026-10-17 20:00 UTC.
- Context
- Coval publishes rolling 30-day results from daily repeated runs on fixed public audio or prompt sets. These rows are the later timestamped run visible in the captured page, not a copy of the earlier run in the input. Exact API/checkpoint revision, request parameters, and confidence intervals are not published; displayed WER is rounded to one decimal and is not a measure of naturalness.
Short-form English WER, Mean 3 sets (Show measurement, test conditions, and source)
- Source value
- 5.04
- Score
- 5.04
- Metric
- normalized word error rate (WER), mean of 3 short-form English sets
- Unit
- %
- Benchmark version
- AssemblyAI Table 7; runs 2026-08-18 to 2026-09-22
- Category
- speech-to-text
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- AssemblyAI speech benchmark
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. Exact source row label: Google Gemini 3.5 Transcribe Live. Table 7 reports normalized word error rate (WER), mean of 3 short-form English sets; displayed components are Common Voice 8.22%, LibriSpeech clean 2.07%, LibriSpeech other 4.83%. Run window: 2026-08-18 to 2026-09-22; the page says one row per model, most recent run. The page does not identify the exact serving route or immutable model revision, per-model N, or decoding settings.
- Context
- AssemblyAI publishes comparative speech benchmarks and is itself a speech vendor; these are vendor-reported, not neutral third-party measurements. The page reports one row per model, the most recent run in the stated window. It does not publish a per-model sample count or confidence interval, the exact competitor endpoint/API route, or an immutable serving revision.
Accented English WER, Mean 4 sets (Show measurement, test conditions, and source)
- Source value
- 9.39
- Score
- 9.39
- Metric
- normalized word error rate (WER), mean of 4 accented-English sets
- Unit
- %
- Benchmark version
- AssemblyAI Table 8; runs 2026-08-18 to 2026-09-22
- Category
- speech-to-text
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- AssemblyAI speech benchmark
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. Exact source row label: Google Gemini 3.5 Transcribe Live. Table 8 reports normalized word error rate (WER), mean of 4 accented-English sets; displayed components are British 6.30%, French-accented 11.86%, Spanish-accented 12.82%, Indian-accented 6.59%. Run window: 2026-08-18 to 2026-09-22; the page says one row per model, most recent run. The page does not identify the exact serving route or immutable model revision, per-model N, or decoding settings.
- Context
- AssemblyAI publishes comparative speech benchmarks and is itself a speech vendor; these are vendor-reported, not neutral third-party measurements. The page reports one row per model, the most recent run in the stated window. It does not publish a per-model sample count or confidence interval, the exact competitor endpoint/API route, or an immutable serving revision.
Page 1 of 16
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-1, pyannote: 73
Nemotron-3-Diarization, NVIDIA: 57
Gemini 3.5 Transcribe, Google: 52
Confucius4-R2T2, NetEase Youdao: 52
Granite Speech 5.0 470M TurboCTC, IBM: 52
Kokoro-82M, hexgrad: 51
Nova-3, Deepgram: 51
Fish Audio S2 Pro, Fish Audio: 51
NVIDIA Nemotron 3.5 Transcribe, NVIDIA: 50
Granite Speech 5.0 470M TurboCTC NC, IBM: 50
Orukeet, Oruk: 49
Lyria 3.5, Google: 47
Gemini 3.5 Live Translate, Google: 46
Phonon-2, Fermion Research: 44
Aura-2, Deepgram: 44
Gemini 3.1 Flash TTS, Google: 44
MiniMax-Music3, MiniMax: 41
Audio8-ASR-Infinite, Edge0 AI: 41
GPT-Realtime 2.1, OpenAI: 41
Gemini 3.5 Transcribe Live, Google: 35
Gemini 3.5 Transcribe Live
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 5, 2026
Gemini 3.5 Transcribe LiveGPT-Realtime TranslateGPT-4o Mini TTSGPT-Realtime-WhisperFlux
All specifications
Architecture
Input and output
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
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
API
Further details
Sources and data date
Show all 29 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.