OpenAIAudio modelActive
GPT-Live-Transcribe
PopularityRank 287 of 631
39out of 100same as last month
based on two data groups, mostly search
Summary
GPT-Live-Transcribe is the streaming option for ongoing speech. It delivers low-latency transcript deltas from an audio stream and lets you tune latency. Unstructured context, keyword hints, and language hints help with domain terms and code-switching. Unlike GPT-Transcribe, the model is built around live speech rather than completed recordings.
GPT-Live-Transcribe returns text only even though input arrives as an audio stream. OpenAI bills it by audio minute; function calling and a ready-made voice agent are outside this transcription profile.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 24 published benchmark results
Page 1 of 2
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.
VoiceCodeBench: CTEM (Show measurement, test conditions, and source)
- Source value
- 91.363
- Score
- 91.363
- Metric
- canonical token/entity match
- Unit
- %
- Benchmark version
- 1
- Category
- general
- Task
- ctem
- Task label
- CTEM
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Vals AI
- Status
- active
- Retrieved at
- 2026-11-06
- Methodology
- Source model ID: openai/gpt-live-transcribe. Source model name: GPT-Live-Transcribe. Provider reported by source: OpenAI. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-09-24. Latency in seconds: 70.342. Standard error: 0. Cost per task in USD: 0.018995. Source ID: vals-ai. Evidence capture ID: vals-voice-code-bench-2026-11-06-7753d162bfe3. Source payload hash: 7753d162bfe3fa3470d586da3917bf8ddd96f14e10f9aa8f03dd517b8d5f003b. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-voice-code-bench:1:ctem
VoiceCodeBench: Overall (Show measurement, test conditions, and source)
- Source value
- 67.667
- Score
- 67.667
- Metric
- task success rate
- Unit
- %
- Benchmark version
- 1
- Category
- general
- Task
- overall
- Task label
- Overall
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Vals AI
- Status
- active
- Retrieved at
- 2026-11-06
- Methodology
- Source model ID: openai/gpt-live-transcribe. Source model name: GPT-Live-Transcribe. Provider reported by source: OpenAI. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-09-24. Latency in seconds: 70.342. Standard error: 0. Cost per task in USD: 0.018995. Source ID: vals-ai. Evidence capture ID: vals-voice-code-bench-2026-11-06-7753d162bfe3. Source payload hash: 7753d162bfe3fa3470d586da3917bf8ddd96f14e10f9aa8f03dd517b8d5f003b. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-voice-code-bench:1:overall
Artificial Analysis AA-WER Streaming (final transcription) (Show measurement, test conditions, and source)
- Source value
- 3.9177
- Score
- 3.9177
- 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.
A5Sv2 ASR benchmark, Mega-ASR fixed robustness selection (Show measurement, test conditions, and source)
- Source value
- 27.812196
- Score
- 27.812196
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- A5Sv2 pinned release commit 64284911a02b38ce07e1d184c56b9c1335f20b57
- Category
- audio
- Direction
- Lower is better
- Sample
- 32928
- Source type
- independent-evaluator
- Evaluator
- AirCaps A5Sv2 ASR benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact API model ID gpt-live-transcribe; English, xhigh reasoning effort, manual commit, 100 ms PCM chunks. Final transcripts only, one saved run per system and corpus, frozen word-onset v2 references. Mega-ASR uses a fixed sample of 1,250 utterances from Voices-in-the-Wild-2M training splits, not its standard test set.
- Context
- The benchmark reports one saved run, without repeated trials or a confidence interval. The API model ID is not bound to an immutable serving build. AirCaps says its own A5Sv2 checkpoint was not optimized on the exact Mega-ASR items, but cannot determine whether third-party systems saw overlapping upstream data.
A5Sv2 ASR benchmark, AMI scenario-only evaluation meetings (Show measurement, test conditions, and source)
- Source value
- 30.159742
- Score
- 30.159742
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- A5Sv2 pinned release commit 64284911a02b38ce07e1d184c56b9c1335f20b57
- Category
- audio
- Direction
- Lower is better
- Sample
- 32928
- Source type
- independent-evaluator
- Evaluator
- AirCaps A5Sv2 ASR benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact API model ID gpt-live-transcribe; English, xhigh reasoning effort, manual commit, 100 ms PCM chunks. Final transcripts only, one saved run per system and corpus, frozen word-onset v2 references. AMI uses seven scenario-only unseen-evaluation meetings and Array1-01 audio.
- Context
- The benchmark reports one saved run, without repeated trials or a confidence interval. The API model ID is not bound to an immutable serving build. AirCaps says its own A5Sv2 checkpoint was not optimized on the exact Mega-ASR items, but cannot determine whether third-party systems saw overlapping upstream data.
A5Sv2 ASR benchmark, DiPCo evaluation and development meetings (Show measurement, test conditions, and source)
- Source value
- 39.861635
- Score
- 39.861635
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- A5Sv2 pinned release commit 64284911a02b38ce07e1d184c56b9c1335f20b57
- Category
- audio
- Direction
- Lower is better
- Sample
- 33679
- Source type
- independent-evaluator
- Evaluator
- AirCaps A5Sv2 ASR benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact API model ID gpt-live-transcribe; English, xhigh reasoning effort, manual commit, 100 ms PCM chunks. Final transcripts only, one saved run per system and corpus, frozen word-onset v2 references. DiPCo uses five evaluation meetings plus one development meeting, U01.CH1 audio, and fixed +5.1 dB gain.
- Context
- The benchmark reports one saved run, without repeated trials or a confidence interval. The API model ID is not bound to an immutable serving build. AirCaps says its own A5Sv2 checkpoint was not optimized on the exact Mega-ASR items, but cannot determine whether third-party systems saw overlapping upstream data.
A5Sv2 ASR benchmark, NOTSOFAR eval-small meetings (Show measurement, test conditions, and source)
- Source value
- 40.845575
- Score
- 40.845575
- Metric
- word error rate (WER)
- Unit
- %
- Benchmark version
- A5Sv2 pinned release commit 64284911a02b38ce07e1d184c56b9c1335f20b57
- Category
- audio
- Direction
- Lower is better
- Sample
- 33492
- Source type
- independent-evaluator
- Evaluator
- AirCaps A5Sv2 ASR benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact API model ID gpt-live-transcribe; English, xhigh reasoning effort, manual commit, 100 ms PCM chunks. Final transcripts only, one saved run per system and corpus, frozen word-onset v2 references. NOTSOFAR uses 22 eval-small meetings and sc_meetup_0/ch0.wav audio.
- Context
- The benchmark reports one saved run, without repeated trials or a confidence interval. The API model ID is not bound to an immutable serving build. AirCaps says its own A5Sv2 checkpoint was not optimized on the exact Mega-ASR items, but cannot determine whether third-party systems saw overlapping upstream data.
A5Sv2 ASR benchmark, unweighted macro mean across four corpora (Show measurement, test conditions, and source)
- Source value
- 34.669787
- Score
- 34.669787
- Metric
- macro mean of corpus-level word error rates (WER)
- Unit
- %
- Benchmark version
- A5Sv2 pinned release commit 64284911a02b38ce07e1d184c56b9c1335f20b57
- Category
- audio
- Direction
- Lower is better
- Sample
- 4
- Source type
- independent-evaluator
- Evaluator
- AirCaps A5Sv2 ASR benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Arithmetic mean of the four corpus WERs for exact API model ID gpt-live-transcribe; English, xhigh reasoning effort, manual commit, 100 ms PCM chunks. One saved run per system and corpus; fixed corpus selections and frozen word-onset v2 references. Mega-ASR is a separate developer-created robustness diagnostic.
- Context
- This aggregate weights the four corpus WERs equally; it is not the pooled word-level WER. No confidence interval or immutable API serving revision is published.
A5Sv2 ASR benchmark, pooled WER across four corpora (Show measurement, test conditions, and source)
- Source value
- 34.725281
- Score
- 34.725281
- Metric
- pooled word error rate (WER)
- Unit
- %
- Benchmark version
- A5Sv2 pinned release commit 64284911a02b38ce07e1d184c56b9c1335f20b57
- Category
- audio
- Direction
- Lower is better
- Sample
- 133027
- Source type
- independent-evaluator
- Evaluator
- AirCaps A5Sv2 ASR benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Pooled errors divided by all 133,027 normalized reference words across the four corpora for exact API model ID gpt-live-transcribe; English, xhigh reasoning effort, manual commit, 100 ms PCM chunks. One saved run per system and corpus; frozen word-onset v2 references.
- Context
- This token-pooled aggregate weights corpora by their reference-word counts; it is not the unweighted macro mean. No confidence interval or immutable API serving revision is published.
OpenAI Context Aware ASR semantic accuracy without free-form context (Show measurement, test conditions, and source)
- Source value
- 38.5
- Score
- 38.5
- Metric
- semantic accuracy
- Unit
- %
- Benchmark version
- Community forum post dated 2026-07-29; benchmark revision not stated
- Category
- audio
- Direction
- Higher is better
- Source type
- community-reported
- Evaluator
- OpenAI Developer Community forum post by VeitB
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Community-reported result. A captured OpenAI Developer Community forum post by VeitB states that GPT-Live-Transcribe scored 38.5% semantic accuracy on the Context Aware Automatic Speech Recognition benchmark without free-form context. The post does not publish sample count, dataset split, scoring protocol, confidence interval or serving revision.
- Context
- The captured source identifies the author as VeitB but does not substantiate an OpenAI affiliation or verify this as a vendor-reported measurement. Treat it as an unverified community-post claim, not a confirmed OpenAI result. N, confidence interval, benchmark revision and immutable serving snapshot are not published. This is one context configuration, not an absolute transcription error rate.
OpenAI Context Aware ASR semantic accuracy with free-form context (Show measurement, test conditions, and source)
- Source value
- 44.6
- Score
- 44.6
- Metric
- semantic accuracy
- Unit
- %
- Benchmark version
- Community forum post dated 2026-07-29; benchmark revision not stated
- Category
- audio
- Direction
- Higher is better
- Source type
- community-reported
- Evaluator
- OpenAI Developer Community forum post by VeitB
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Community-reported result. A captured OpenAI Developer Community forum post by VeitB states that GPT-Live-Transcribe scored 44.6% semantic accuracy on the Context Aware Automatic Speech Recognition benchmark with free-form context. The post does not publish sample count, dataset split, scoring protocol, confidence interval or serving revision. This is a separate prompt/context configuration from the 38.5% no-context result.
- Context
- The captured source identifies the author as VeitB but does not substantiate an OpenAI affiliation or verify this as a vendor-reported measurement. Treat it as an unverified community-post claim, not a confirmed OpenAI result. N, confidence interval, benchmark revision and immutable serving snapshot are not published. This is one context configuration, not an absolute transcription error rate.
OpenAI Common Voice transcription error rate across 22 languages (Show measurement, test conditions, and source)
- Source value
- 19.7
- Score
- 19.7
- Metric
- transcription error rate, metric formula not stated
- Unit
- %
- Benchmark version
- Community forum post dated 2026-07-29; benchmark revision not stated
- Category
- audio
- Direction
- Lower is better
- Source type
- community-reported
- Evaluator
- OpenAI Developer Community forum post by VeitB
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Community-reported result. A captured OpenAI Developer Community forum post by VeitB states that GPT-Live-Transcribe scored a 19.70% transcription error rate across 22 languages on Common Voice. The post does not specify the Common Voice release, split, language aggregation, sample count, formula, confidence interval or serving revision.
- Context
- The captured source identifies the author as VeitB but does not substantiate an OpenAI affiliation or verify this as a vendor-reported measurement. Treat it as an unverified community-post claim, not a confirmed OpenAI result. The post does not state whether this is WER or another transcription error metric, nor provide N, confidence interval or benchmark revision. It is not a directly comparable task score.
Page 1 of 2
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 45 models
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 7, 2026
GPT-Live-TranscribeMaya 2 Native 40Eleven v3 39Scribe v2 39Sonic 3.6 39
Method changes in the timeline, a jump at these points is no market event:
- 1November 1, 2026: Value only with search or usage data, 30-day attention
What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.
GPT-Live-Transcribe compared
Similar models
More models of the same kind as GPT-Live-Transcribe, chosen by model line, developer, size, price and release date.
- GPT-TranscribeOpenAI · same developer
- GPT-Realtime 2.1 MiniOpenAI · same developer
- GPT-Live 1OpenAI · same developer
- GPT-Realtime 2.1OpenAI · same developer
- GPT-Realtime TranslateOpenAI · same developer
- GPT-Realtime-WhisperOpenAI · same developer
- MAI-Voice-2-FlashMicrosoft · released around the same time
- Cartesia Ink-2Cartesia · released around the same time
All specifications
Architecture
Input and output
- Capabilities
- Tunable latency, unstructured context, keyword hints, and language hints
- Source: Capabilities
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
Prices
- Live transcription price
- $0.017 per audio minuteSource
Sources and data date
Show all 24 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
| Evaluator | Test series | Composite indices | Recorded active measurements |
|---|---|---|---|
| Vals AI | 1 | 0 | 3 |
Without vendor claims and historical or disputed results; those stay labelled in the single values.