Skip to main content

Edge0 AIAudio modelOpen weights

Audio8-ASR-Infinite

PopularityRank 20 of 24 audio models

22out of 100new calculation

based on two data groups, mostly search, range 12 to 31

Summary

Audio8-ASR-Infinite is an open streaming specialist for Chinese and English speech recognition. Its card targets ongoing, near-unbounded audio streams and returns transcript text rather than audio.

The model can run locally from the published Edge0-AI checkpoint. Apache-2.0 covers the weights and source code. This is a preview release, with the formal launch of semantic perception features still pending according to the project.

GraduallyBench Audio

Audio8-ASR-Infinite

Partially tested
  • The test coverage is incomplete.
  • The model’s language coverage was not fully checked.
  • Only the indicated language subset was tested.
Show test details

Speech recognition

Local · English

20 valid outputs from 20 attempted and 36 planned test slots.

  • The test process stopped unexpectedly.
  • Some cases were answered after a retry.
Word error rate (WER)
2.86 %(14 / 490 reference words)English
Silence cases with words in the final transcript
0 %(0 / 4 assessable silence cases)EnglishAfter Unicode normalization, word segments containing letters or digits count. Punctuation-only output does not count.

Speech recognition

Local · German

4 valid outputs from 4 attempted and 4 planned test slots.

  • The English language token was forced for these German recordings. German is outside the documented Chinese and English coverage.
Word error rate (WER)
69.72 %(76 / 109 reference words)German
Character error rate (CER)
29.83 %(196 / 657 reference characters)German

WER compares words and CER compares characters with the reference. The speech output was re-transcribed with GPT Transcribe. Its exact backend revision is unconfirmed. Read speech and digital silence do not cover spontaneous speech or real background noise. Error rates can exceed 100 %. Where used, the test references were not independently reviewed by humans. Test coverage and methods differ by model. Each metric stands on its own; there is no combined overall score across tests. Data compiled on 11/10/2026. How we test · Download all Decision and Audio results as JSON

All benchmark results

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

Show all 12 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.

LibriSpeech test-clean (40 utterances) WER (Show measurement, test conditions, and source)
Source value
4.27
Metric
word error rate (WER)
Unit
%
Direction
No direction published
Sample
40
Source type
community-reported
Evaluator
Masterx Audio8-ASR-Infinite ONNX model card / WinSTT Rust runtime
Status
active
Retrieved at
2026-10-18
Methodology
Community-reported result. Audio8-ASR-Infinite ONNX int8; WinSTT Rust runtime on CPU; 40 LibriSpeech test-clean utterances (308 s), streamed as a continuous recording of about five minutes. English scoring lower-cased with punctuation stripped. Exact base checkpoint: Edge0/Audio8-ASR-Infinite revision 7476824bc222e4ad509d286e8cae8b8d3f371129.
Source Masterx Audio8-ASR-Infinite ONNX model card / WinSTT Rust runtime
FLEURS cmn_hans_cn (25 utterances) CER (Show measurement, test conditions, and source)
Source value
11.25
Metric
character error rate (CER)
Unit
%
Direction
No direction published
Sample
25
Source type
community-reported
Evaluator
Masterx Audio8-ASR-Infinite ONNX model card / WinSTT Rust runtime
Status
active
Retrieved at
2026-10-18
Methodology
Community-reported result. Audio8-ASR-Infinite ONNX int8; WinSTT Rust runtime on CPU; 25 FLEURS cmn_hans_cn utterances (290 s), streamed as a continuous recording of about five minutes. Chinese scoring uses NFKC normalization and strips punctuation and whitespace; digits and Chinese numerals are not unified. Exact base checkpoint: Edge0/Audio8-ASR-Infinite revision 7476824bc222e4ad509d286e8cae8b8d3f371129.
Source Masterx Audio8-ASR-Infinite ONNX model card / WinSTT Rust runtime
Transcribing English Audio We Made (Cloud GPU) (Show measurement, test conditions, and source)
Source value
0
Metric
word error rate (WER)
Unit
%
Direction
No direction published
Sample
3
Source type
independent-evaluator
Evaluator
Local Model Watch
Status
active
Retrieved at
2026-10-19
Methodology
Independent evaluation. English-only transcription of three fixed English sentences spoken using the en_US-ljspeech-medium TTS voice; one first output per clip, with nothing selected or edited. Base model Edge0/Audio8-ASR-Infinite at revision 7476824bc222e4ad509d286e8cae8b8d3f371129, BF16, publisher's streaming decoder at commit c8ba8eea829be0339e8d7757f8ca52dac06e1e32, NVIDIA L4 (22 GB) on Modal, Transformers 5.13.0, PyTorch 2.13.0+cu130, greedy decoding, 80 ms clock and 480 ms delay. WER counted after lowercasing and punctuation removal; numerals are not converted. Measured 2026-10-01 (JST).
Context
This is a three-clip synthetic-voice English test, not a public-corpus benchmark. The source explicitly says it does not establish accuracy on real recordings, accents, or noise; it also does not measure real-time latency.
Source Local Model Watch
CER (Show measurement, test conditions, and source)
Source value
1.75
Metric
character error rate
Unit
%
Benchmark version
source version not stated
Category
audio
Direction
Lower is better
Source type
vendor-reported
Evaluator
Audio8-ASR-Infinite model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Card benchmark
Context
Source-specific observation; it does not establish a compatible cross-model cohort.
Source Audio8-ASR-Infinite model card
WER (Show measurement, test conditions, and source)
Source value
3.042
Metric
word error rate
Unit
%
Benchmark version
source version not stated
Category
audio
Direction
Lower is better
Source type
vendor-reported
Evaluator
Audio8-ASR-Infinite model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Card benchmark
Context
Source-specific observation; it does not establish a compatible cross-model cohort.
Source Audio8-ASR-Infinite model card
WER (Show measurement, test conditions, and source)
Source value
6.808
Metric
word error rate
Unit
%
Benchmark version
source version not stated
Category
audio
Direction
Lower is better
Source type
vendor-reported
Evaluator
Audio8-ASR-Infinite model card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Card benchmark
Context
Source-specific observation; it does not establish a compatible cross-model cohort.
Source Audio8-ASR-Infinite model card
Audio8-ASR-Infinite model-card AISHELL-4/test CER (Show measurement, test conditions, and source)
Source value
2.893
Metric
CER
Unit
%
Benchmark version
Audio8-ASR-Infinite preview card revision 7476824; AISHELL-4/test
Category
audio
Direction
Lower is better
Source type
vendor-reported
Evaluator
Edge0 Audio8-ASR-Infinite model card
Status
active
Retrieved at
2026-10-16
Methodology
Vendor-reported result. Preview ASR checkpoint; AISHELL-4/test character error rate. 80 ms audio clock, 480 ms transcription-delay setting (six delay tokens), greedy decoding with EOS suppressed. Error rates are percent.
Context
The card does not publish evaluation-set sample counts or confidence intervals. This is AISHELL-4 CER only; the heterogeneous WER/CER average in the same table is excluded.
Source Edge0 Audio8-ASR-Infinite model card
Community reserved-11 evaluation, English WER, original BF16 CUDA (Show measurement, test conditions, and source)
Source value
0.1932418163
Metric
Word error rate (source error_rate; lower is better)
Unit
ratio
Benchmark version
Community custom reserved evaluation; history shows current-main commit c49eca9 (short SHA; immutable full revision not resolved)
Category
custom English speech recognition
Direction
Lower is better
Source type
community-reported
Evaluator
Audio8-ASR-Infinite community reserved-11 paired evaluation artifact
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. The artifact labels this result original BF16 CUDA and the associated card identifies the exact base as Edge0/Audio8-ASR-Infinite revision b4413de154ed6bdef0a4011028b1ebd12aca8152. Custom reserved evaluation: 183 word errors / 947 English reference words, source error_rate 0.1932418162618796. The card describes 11 recordings overall; per-language clip count and evaluation date are not stated.
Context
Community-reported custom material, not a public standardized corpus or a broad quality ranking. The result artifact is hosted beside a compressed derivative, but the reported row is explicitly the original BF16 CUDA control and the card pins its upstream base revision. No confidence interval is published.
Source Audio8-ASR-Infinite community reserved-11 paired evaluation artifact
Community reserved-11 evaluation, Chinese CER, original BF16 CUDA (Show measurement, test conditions, and source)
Source value
0.0084745763
Metric
Character error rate (source error_rate; lower is better)
Unit
ratio
Benchmark version
Community custom reserved evaluation; history shows current-main commit c49eca9 (short SHA; immutable full revision not resolved)
Category
custom Chinese speech recognition
Direction
Lower is better
Source type
community-reported
Evaluator
Audio8-ASR-Infinite community reserved-11 paired evaluation artifact
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. The artifact labels this result original BF16 CUDA and the associated card identifies the exact base as Edge0/Audio8-ASR-Infinite revision b4413de154ed6bdef0a4011028b1ebd12aca8152. Custom reserved evaluation: 1 character error / 118 Chinese reference characters, source error_rate 0.00847457627118644. The card describes 11 recordings overall; per-language clip count and evaluation date are not stated.
Context
Community-reported custom material, not a public standardized corpus or a broad quality ranking. The result artifact is hosted beside a compressed derivative, but the reported row is explicitly the original BF16 CUDA control and the card pins its upstream base revision. No confidence interval is published.
Source Audio8-ASR-Infinite community reserved-11 paired evaluation artifact
Local Model Watch, synthetic English clip 1, transcription time (Show measurement, test conditions, and source)
Source value
6.49
Metric
One-pass transcription elapsed time for one fixed synthetic English utterance, not real-time latency
Unit
seconds
Direction
Lower is better
Sample
1
Source type
independent-evaluator
Evaluator
Local Model Watch
Status
active
Retrieved at
2026-10-22
Methodology
Independent evaluation. One fixed synthetic English TTS utterance was transcribed once using BF16 weights at Hugging Face revision 7476824bc222e4ad509d286e8cae8b8d3f371129 and the publisher's streaming decoder at commit c8ba8eea829be0339e8d7757f8ca52dac06e1e32. The source reports an NVIDIA L4 on Modal, transformers 5.13.0, PyTorch 2.13.0+cu130, greedy decoding, and a run on 2026-10-01 JST. Clip: 4.7-second fixed English TTS clip. The source says the decoder replays the stream once and does not measure real-time latency.
Context
A single synthetic clip is not a general speech-quality test. The source explicitly says this replay does not measure real-time latency.
Source Local Model Watch
Local Model Watch, synthetic English clip 2, transcription time (Show measurement, test conditions, and source)
Source value
6.5
Metric
One-pass transcription elapsed time for one fixed synthetic English utterance, not real-time latency
Unit
seconds
Direction
Lower is better
Sample
1
Source type
independent-evaluator
Evaluator
Local Model Watch
Status
active
Retrieved at
2026-10-22
Methodology
Independent evaluation. One fixed synthetic English TTS utterance was transcribed once using BF16 weights at Hugging Face revision 7476824bc222e4ad509d286e8cae8b8d3f371129 and the publisher's streaming decoder at commit c8ba8eea829be0339e8d7757f8ca52dac06e1e32. The source reports an NVIDIA L4 on Modal, transformers 5.13.0, PyTorch 2.13.0+cu130, greedy decoding, and a run on 2026-10-01 JST. Clip: 4.7-second fixed English TTS clip. The source says the decoder replays the stream once and does not measure real-time latency.
Context
A single synthetic clip is not a general speech-quality test. The source explicitly says this replay does not measure real-time latency.
Source Local Model Watch
Local Model Watch, synthetic English clip 3, transcription time (Show measurement, test conditions, and source)
Source value
5.36
Metric
One-pass transcription elapsed time for one fixed synthetic English utterance, not real-time latency
Unit
seconds
Direction
Lower is better
Sample
1
Source type
independent-evaluator
Evaluator
Local Model Watch
Status
active
Retrieved at
2026-10-22
Methodology
Independent evaluation. One fixed synthetic English TTS utterance was transcribed once using BF16 weights at Hugging Face revision 7476824bc222e4ad509d286e8cae8b8d3f371129 and the publisher's streaming decoder at commit c8ba8eea829be0339e8d7757f8ca52dac06e1e32. The source reports an NVIDIA L4 on Modal, transformers 5.13.0, PyTorch 2.13.0+cu130, greedy decoding, and a run on 2026-10-01 JST. Clip: 3.61-second fixed English TTS clip. The source says the decoder replays the stream once and does not measure real-time latency.
Context
A single synthetic clip is not a general speech-quality test. The source explicitly says this replay does not measure real-time latency.
Source Local Model Watch

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 54 models

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

Over time

Weekly values from October 18, 2026 to November 20, 2026

Audio8-ASR-InfiniteEleven Flash v2.5 22Index-Echo-S2ST-2B 22Index-Echo-S2TT-2B 21Index-Echo-S2ST-9B 21

Method changes in the timeline, a jump at these points is no market event:

  1. 1November 1, 2026: Value only with search or usage data, 30-day attention
  2. 2November 1, 2026: Method version 4 (common scale, shrinkage)
  3. 3November 9, 2026: New topics measurement V20 (daily quorum, short forms)

What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.

All specifications

Architecture

Model class
Streaming speech recognition
Source: Model class
Checkpoint
Edge0/Audio8-ASR-Infinite
Source: Checkpoint

Input and output

Input
Chinese and English audio
Source: Input
Output
Streaming transcript
Source: 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

Access
Local weights
Source: Access
License
Apache-2.0
Source: License
Lifecycle
Preview release of open weights; the formal semantic-perception release remains in progress
Source: Lifecycle
Shutdown
None announced
no source

Further details

Streaming
Designed for ongoing and near-unbounded ASR streams
Source: Streaming
Repository
Source code is published in the Edge0-AI repository
Source: Repository

Sources and data date

Show all 12 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.

TypeEvidence and data date
Research statusResearch date October 4, 2026. 5 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not published

evaluation-sample-count

Note

The preview card does not state the evaluated example count for AISHELL-4/test.

Not published

confidence-intervals

Note

The preview card provides no confidence interval for its CER point estimate.

Additional sourceAudio8-ASR-Infinite model card (retrieved October 3, 2026; October 4, 2026; October 18, 2026) · Audio8-ASR-Infinite model card · Editorial description reviewed October 3, 2026
Additional sourceAudio8-ASR-Infinite repository (retrieved October 3, 2026) · Audio8-ASR-Infinite repository · Editorial description reviewed October 3, 2026
Additional sourceEdge0 Audio8-ASR-Infinite model card (retrieved October 16, 2026)
Additional sourceMasterx Audio8-ASR-Infinite ONNX model card / WinSTT Rust runtime (retrieved October 18, 2026)
Additional sourceLocal Model Watch (retrieved October 19, 2026; October 22, 2026)
Additional sourceAudio8-ASR-Infinite community reserved-11 paired evaluation artifact (retrieved October 19, 2026)
Additional sourceOpenRouter: Popularity index (derived), CC BY 4.0
Additional sourceVercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional sourceOllama: cumulative downloads for the popularity index
Additional sourceReplicate: cumulative runs for the popularity index
Additional sourceGoogle Trending Now: limited feed selection for Germany and the US
Additional sourceCreative Commons Attribution 4.0