IBMAudio modelOpen weights
Granite Speech 5.0 470M TurboCTC
PopularityRank 89 of 875
52out of 100same as last month
Summary
Granite Speech 5.0 470M TurboCTC is a 470-million-parameter model for English speech recognition. Its Conformer architecture uses non-autoregressive CTC decoding and targets edge devices, laptops, and mobile hardware.
The checkpoint is available locally under Apache-2.0. Its role is audio to text, not audio understanding with a spoken response.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 29 published benchmark results
Page 1 of 3
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.
FFASR average WER: Near Field (Show measurement, test conditions, and source)
- Source value
- 4.8
- Score
- 4.8
- Metric
- Average word error rate (WER), FFASR scenario Near Field Speech (close-mic)
- Unit
- %
- Benchmark version
- Official FFASR leaderboard snapshot shown in IBM model card as of 2026-08-25
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech model card, FFASR WER figure
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. FFASR leaderboard WER heatmap embedded in the IBM Granite Speech 5.0 model card pinned at revision 947f59af40db9791170a0628cf0f3f4812d720f1. Figure caption dates official FFASR results to 2026-08-25 and names nine far-field, noisy, reverberant or moving-speaker scenarios. Exact FFASR row label: granite-speech-5.0-470m-turboctc. Do not treat this as the separate NC variant. Exact chart column: Near Field; evaluator scenario description: Near Field Speech (close-mic). The figure shows average WER, but does not show this cell's evaluated audio count or CI.
- Context
- The vendor model card reproduces this as an official FFASR leaderboard snapshot as of 2026-08-25. The actual figure does not publish a cell-level sample count or confidence interval; the current evaluator code confirms a num_samples CSV field exists but the historical CSV was not retrieved. The adjacent one-L4-GPU note applies to RTFx, not WER hardware. The card attributes the benchmark to the Treble Technologies FFASR leaderboard; this record preserves the vendor-published image rather than claiming an independent reproduction.
FFASR average WER: Lab Meas. (Show measurement, test conditions, and source)
- Source value
- 26.2
- Score
- 26.2
- Metric
- Average word error rate (WER), FFASR scenario Lab Measured
- Unit
- %
- Benchmark version
- Official FFASR leaderboard snapshot shown in IBM model card as of 2026-08-25
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech model card, FFASR WER figure
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. FFASR leaderboard WER heatmap embedded in the IBM Granite Speech 5.0 model card pinned at revision 947f59af40db9791170a0628cf0f3f4812d720f1. Figure caption dates official FFASR results to 2026-08-25 and names nine far-field, noisy, reverberant or moving-speaker scenarios. Exact FFASR row label: granite-speech-5.0-470m-turboctc. Do not treat this as the separate NC variant. Exact chart column: Lab Meas.; evaluator scenario description: Lab Measured. The figure shows average WER, but does not show this cell's evaluated audio count or CI.
- Context
- The vendor model card reproduces this as an official FFASR leaderboard snapshot as of 2026-08-25. The actual figure does not publish a cell-level sample count or confidence interval; the current evaluator code confirms a num_samples CSV field exists but the historical CSV was not retrieved. The adjacent one-L4-GPU note applies to RTFx, not WER hardware. The card attributes the benchmark to the Treble Technologies FFASR leaderboard; this record preserves the vendor-published image rather than claiming an independent reproduction.
FFASR average WER: Lab Sim. (Show measurement, test conditions, and source)
- Source value
- 24.9
- Score
- 24.9
- Metric
- Average word error rate (WER), FFASR scenario Lab Simulated
- Unit
- %
- Benchmark version
- Official FFASR leaderboard snapshot shown in IBM model card as of 2026-08-25
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech model card, FFASR WER figure
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. FFASR leaderboard WER heatmap embedded in the IBM Granite Speech 5.0 model card pinned at revision 947f59af40db9791170a0628cf0f3f4812d720f1. Figure caption dates official FFASR results to 2026-08-25 and names nine far-field, noisy, reverberant or moving-speaker scenarios. Exact FFASR row label: granite-speech-5.0-470m-turboctc. Do not treat this as the separate NC variant. Exact chart column: Lab Sim.; evaluator scenario description: Lab Simulated. The figure shows average WER, but does not show this cell's evaluated audio count or CI.
- Context
- The vendor model card reproduces this as an official FFASR leaderboard snapshot as of 2026-08-25. The actual figure does not publish a cell-level sample count or confidence interval; the current evaluator code confirms a num_samples CSV field exists but the historical CSV was not retrieved. The adjacent one-L4-GPU note applies to RTFx, not WER hardware. The card attributes the benchmark to the Treble Technologies FFASR leaderboard; this record preserves the vendor-published image rather than claiming an independent reproduction.
FFASR average WER: High SNR (Show measurement, test conditions, and source)
- Source value
- 10.5
- Score
- 10.5
- Metric
- Average word error rate (WER), FFASR scenario Realistic conditions: High SNR
- Unit
- %
- Benchmark version
- Official FFASR leaderboard snapshot shown in IBM model card as of 2026-08-25
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech model card, FFASR WER figure
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. FFASR leaderboard WER heatmap embedded in the IBM Granite Speech 5.0 model card pinned at revision 947f59af40db9791170a0628cf0f3f4812d720f1. Figure caption dates official FFASR results to 2026-08-25 and names nine far-field, noisy, reverberant or moving-speaker scenarios. Exact FFASR row label: granite-speech-5.0-470m-turboctc. Do not treat this as the separate NC variant. Exact chart column: High SNR; evaluator scenario description: Realistic conditions: High SNR. The figure shows average WER, but does not show this cell's evaluated audio count or CI.
- Context
- The vendor model card reproduces this as an official FFASR leaderboard snapshot as of 2026-08-25. The actual figure does not publish a cell-level sample count or confidence interval; the current evaluator code confirms a num_samples CSV field exists but the historical CSV was not retrieved. The adjacent one-L4-GPU note applies to RTFx, not WER hardware. The card attributes the benchmark to the Treble Technologies FFASR leaderboard; this record preserves the vendor-published image rather than claiming an independent reproduction.
FFASR average WER: Mov. High (Show measurement, test conditions, and source)
- Source value
- 11.5
- Score
- 11.5
- Metric
- Average word error rate (WER), FFASR scenario Moving High SNR
- Unit
- %
- Benchmark version
- Official FFASR leaderboard snapshot shown in IBM model card as of 2026-08-25
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech model card, FFASR WER figure
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. FFASR leaderboard WER heatmap embedded in the IBM Granite Speech 5.0 model card pinned at revision 947f59af40db9791170a0628cf0f3f4812d720f1. Figure caption dates official FFASR results to 2026-08-25 and names nine far-field, noisy, reverberant or moving-speaker scenarios. Exact FFASR row label: granite-speech-5.0-470m-turboctc. Do not treat this as the separate NC variant. Exact chart column: Mov. High; evaluator scenario description: Moving High SNR. The figure shows average WER, but does not show this cell's evaluated audio count or CI.
- Context
- The vendor model card reproduces this as an official FFASR leaderboard snapshot as of 2026-08-25. The actual figure does not publish a cell-level sample count or confidence interval; the current evaluator code confirms a num_samples CSV field exists but the historical CSV was not retrieved. The adjacent one-L4-GPU note applies to RTFx, not WER hardware. The card attributes the benchmark to the Treble Technologies FFASR leaderboard; this record preserves the vendor-published image rather than claiming an independent reproduction.
FFASR average WER: Mid SNR (Show measurement, test conditions, and source)
- Source value
- 19.3
- Score
- 19.3
- Metric
- Average word error rate (WER), FFASR scenario Realistic conditions: Mid SNR
- Unit
- %
- Benchmark version
- Official FFASR leaderboard snapshot shown in IBM model card as of 2026-08-25
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech model card, FFASR WER figure
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. FFASR leaderboard WER heatmap embedded in the IBM Granite Speech 5.0 model card pinned at revision 947f59af40db9791170a0628cf0f3f4812d720f1. Figure caption dates official FFASR results to 2026-08-25 and names nine far-field, noisy, reverberant or moving-speaker scenarios. Exact FFASR row label: granite-speech-5.0-470m-turboctc. Do not treat this as the separate NC variant. Exact chart column: Mid SNR; evaluator scenario description: Realistic conditions: Mid SNR. The figure shows average WER, but does not show this cell's evaluated audio count or CI.
- Context
- The vendor model card reproduces this as an official FFASR leaderboard snapshot as of 2026-08-25. The actual figure does not publish a cell-level sample count or confidence interval; the current evaluator code confirms a num_samples CSV field exists but the historical CSV was not retrieved. The adjacent one-L4-GPU note applies to RTFx, not WER hardware. The card attributes the benchmark to the Treble Technologies FFASR leaderboard; this record preserves the vendor-published image rather than claiming an independent reproduction.
FFASR average WER: Mov. Mid (Show measurement, test conditions, and source)
- Source value
- 21.5
- Score
- 21.5
- Metric
- Average word error rate (WER), FFASR scenario Moving Mid SNR
- Unit
- %
- Benchmark version
- Official FFASR leaderboard snapshot shown in IBM model card as of 2026-08-25
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech model card, FFASR WER figure
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. FFASR leaderboard WER heatmap embedded in the IBM Granite Speech 5.0 model card pinned at revision 947f59af40db9791170a0628cf0f3f4812d720f1. Figure caption dates official FFASR results to 2026-08-25 and names nine far-field, noisy, reverberant or moving-speaker scenarios. Exact FFASR row label: granite-speech-5.0-470m-turboctc. Do not treat this as the separate NC variant. Exact chart column: Mov. Mid; evaluator scenario description: Moving Mid SNR. The figure shows average WER, but does not show this cell's evaluated audio count or CI.
- Context
- The vendor model card reproduces this as an official FFASR leaderboard snapshot as of 2026-08-25. The actual figure does not publish a cell-level sample count or confidence interval; the current evaluator code confirms a num_samples CSV field exists but the historical CSV was not retrieved. The adjacent one-L4-GPU note applies to RTFx, not WER hardware. The card attributes the benchmark to the Treble Technologies FFASR leaderboard; this record preserves the vendor-published image rather than claiming an independent reproduction.
FFASR average WER: Low SNR (Show measurement, test conditions, and source)
- Source value
- 35.7
- Score
- 35.7
- Metric
- Average word error rate (WER), FFASR scenario Realistic conditions: Low SNR
- Unit
- %
- Benchmark version
- Official FFASR leaderboard snapshot shown in IBM model card as of 2026-08-25
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech model card, FFASR WER figure
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. FFASR leaderboard WER heatmap embedded in the IBM Granite Speech 5.0 model card pinned at revision 947f59af40db9791170a0628cf0f3f4812d720f1. Figure caption dates official FFASR results to 2026-08-25 and names nine far-field, noisy, reverberant or moving-speaker scenarios. Exact FFASR row label: granite-speech-5.0-470m-turboctc. Do not treat this as the separate NC variant. Exact chart column: Low SNR; evaluator scenario description: Realistic conditions: Low SNR. The figure shows average WER, but does not show this cell's evaluated audio count or CI.
- Context
- The vendor model card reproduces this as an official FFASR leaderboard snapshot as of 2026-08-25. The actual figure does not publish a cell-level sample count or confidence interval; the current evaluator code confirms a num_samples CSV field exists but the historical CSV was not retrieved. The adjacent one-L4-GPU note applies to RTFx, not WER hardware. The card attributes the benchmark to the Treble Technologies FFASR leaderboard; this record preserves the vendor-published image rather than claiming an independent reproduction.
FFASR average WER: Mov. Low (Show measurement, test conditions, and source)
- Source value
- 37.2
- Score
- 37.2
- Metric
- Average word error rate (WER), FFASR scenario Moving Low SNR
- Unit
- %
- Benchmark version
- Official FFASR leaderboard snapshot shown in IBM model card as of 2026-08-25
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech model card, FFASR WER figure
- Status
- active
- Retrieved at
- 2026-10-18
- Methodology
- Vendor-reported result. FFASR leaderboard WER heatmap embedded in the IBM Granite Speech 5.0 model card pinned at revision 947f59af40db9791170a0628cf0f3f4812d720f1. Figure caption dates official FFASR results to 2026-08-25 and names nine far-field, noisy, reverberant or moving-speaker scenarios. Exact FFASR row label: granite-speech-5.0-470m-turboctc. Do not treat this as the separate NC variant. Exact chart column: Mov. Low; evaluator scenario description: Moving Low SNR. The figure shows average WER, but does not show this cell's evaluated audio count or CI.
- Context
- The vendor model card reproduces this as an official FFASR leaderboard snapshot as of 2026-08-25. The actual figure does not publish a cell-level sample count or confidence interval; the current evaluator code confirms a num_samples CSV field exists but the historical CSV was not retrieved. The adjacent one-L4-GPU note applies to RTFx, not WER hardware. The card attributes the benchmark to the Treble Technologies FFASR leaderboard; this record preserves the vendor-published image rather than claiming an independent reproduction.
WER (Show measurement, test conditions, and source)
- Source value
- 5
- Score
- 5
- Metric
- word error rate
- Unit
- %
- Benchmark version
- source version not stated
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- IBM Granite Speech announcement
- Status
- active
- Retrieved at
- 2026-10-04
- Methodology
- Vendor-reported result. Apache checkpoint; IBM OpenASR table
- Context
- Source-specific observation; it does not establish a compatible cross-model cohort.
Open ASR leaderboard English short-set WER: AMI (Show measurement, test conditions, and source)
- Source value
- 7.72
- Score
- 7.72
- Metric
- WER
- Unit
- %
- Benchmark version
- hf-audio Open ASR English short-set snapshot 2026-09-25 (dataset commit d2c5b384deccdb82834f41aeaffcc618c00efa2f)
- Category
- audio
- Direction
- Lower is better
- Source type
- official-leaderboard
- Evaluator
- Hugging Face hf-audio Open ASR leaderboard frozen results CSV
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Official benchmark leaderboard. Exact leaderboard checkpoint ibm-granite/granite-speech-5.0-470m-turboctc (Apache-2.0 distribution identity), English short-set WER result for AMI from the frozen 2026-09-25 snapshot. The CSV does not report an evaluated utterance count for this exact checkpoint.
- Context
- The CSV does not state the evaluated utterance count for this exact Granite checkpoint. Counts in the separate Phonon card apply to those Phonon rows and cannot be transferred to Granite; the card gives different Phonon-1 loader counts for AMI and GigaSpeech. No confidence interval is published.
Open ASR leaderboard English short-set WER: Earnings-22 (Show measurement, test conditions, and source)
- Source value
- 6.81
- Score
- 6.81
- Metric
- WER
- Unit
- %
- Benchmark version
- hf-audio Open ASR English short-set snapshot 2026-09-25 (dataset commit d2c5b384deccdb82834f41aeaffcc618c00efa2f)
- Category
- audio
- Direction
- Lower is better
- Source type
- official-leaderboard
- Evaluator
- Hugging Face hf-audio Open ASR leaderboard frozen results CSV
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Official benchmark leaderboard. Exact leaderboard checkpoint ibm-granite/granite-speech-5.0-470m-turboctc (Apache-2.0 distribution identity), English short-set WER result for Earnings-22 from the frozen 2026-09-25 snapshot. The CSV does not report an evaluated utterance count for this exact checkpoint.
- Context
- The leaderboard describes six Earnings-22 sessions scored in chunks but gives no exact denominator for these Granite checkpoints. No N or confidence interval is inferred.
Page 1 of 3
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
Eleven v3, ElevenLabs: 39
Granite Speech 5.0 470M TurboCTC
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 5, 2026
Granite Speech 5.0 470M TurboCTCConfucius4-R2T2Gemini 3.5 TranscribeKokoro-82MNova-3
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
Further details
Sources and data date
Show all 22 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.