Update model weights

.eval_results/open_asr_leaderboard.yaml

@@ -0,0 +1,89 @@
+- dataset:
+    id: hf-audio/open-asr-leaderboard
+    task_id: mean_wer
+  value: 8.91
+  date: '2026-04-14'
+  source:
+    url: https://huggingface.co/hf-audio
+    name: open-asr-leaderboard
+    user: hf-audio
+
+- dataset:
+    id: hf-audio/open-asr-leaderboard
+    task_id: ami_wer
+  value: 15.87
+  date: '2026-04-14'
+  source:
+    url: https://huggingface.co/hf-audio
+    name: open-asr-leaderboard
+    user: hf-audio
+
+- dataset:
+    id: hf-audio/open-asr-leaderboard
+    task_id: earnings22_wer
+  value: 12.09
+  date: '2026-04-14'
+  source:
+    url: https://huggingface.co/hf-audio
+    name: open-asr-leaderboard
+    user: hf-audio
+
+- dataset:
+    id: hf-audio/open-asr-leaderboard
+    task_id: gigaspeech_wer
+  value: 12.56
+  date: '2026-04-14'
+  source:
+    url: https://huggingface.co/hf-audio
+    name: open-asr-leaderboard
+    user: hf-audio
+
+- dataset:
+    id: hf-audio/open-asr-leaderboard
+    task_id: librispeech_clean_wer
+  value: 3.84
+  date: '2026-04-14'
+  source:
+    url: https://huggingface.co/hf-audio
+    name: open-asr-leaderboard
+    user: hf-audio
+
+- dataset:
+    id: hf-audio/open-asr-leaderboard
+    task_id: librispeech_other_wer
+  value: 9.35
+  date: '2026-04-14'
+  source:
+    url: https://huggingface.co/hf-audio
+    name: open-asr-leaderboard
+    user: hf-audio
+
+- dataset:
+    id: hf-audio/open-asr-leaderboard
+    task_id: spgispeech_wer
+  value: 3.10
+  date: '2026-04-14'
+  source:
+    url: https://huggingface.co/hf-audio
+    name: open-asr-leaderboard
+    user: hf-audio
+
+- dataset:
+    id: hf-audio/open-asr-leaderboard
+    task_id: tedlium_wer
+  value: 5.71
+  date: '2026-04-14'
+  source:
+    url: https://huggingface.co/hf-audio
+    name: open-asr-leaderboard
+    user: hf-audio
+
+- dataset:
+    id: hf-audio/open-asr-leaderboard
+    task_id: voxpopuli_wer
+  value: 8.73
+  date: '2026-04-14'
+  source:
+    url: https://huggingface.co/hf-audio
+    name: open-asr-leaderboard
+    user: hf-audio

.gitattributes

@@ -19,6 +19,7 @@
 *.pb filter=lfs diff=lfs merge=lfs -text
 *.pickle filter=lfs diff=lfs merge=lfs -text
 *.pkl filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
 *.pt filter=lfs diff=lfs merge=lfs -text
 *.pth filter=lfs diff=lfs merge=lfs -text
 *.rar filter=lfs diff=lfs merge=lfs -text

README.md

@@ -13,7 +13,7 @@ library_name: transformers
 
 # ABR's niagara-38m-batch.en SSM
 
-The niagara-38m-batch.en model is a [State Space Model](https://huggingface.co/blog/lbourdois/get-on-the-ssm-train) (SSM) with attention that performs automatic speech recognition (ASR), trained and released by [Applied Brain Research](http://www.appliedbrainresearch.com/) (ABR). This model contains \~38m parameters, transcribes speech in English, and was trained on 15k hours of speech data (competitors use about 200k hours). This is a batch model variant designed for direct comparison with other models on the Open ASR Leaderboard. SSMs are an ideal solution for streaming contexts, but to provide a more direct comparison with other models in the leaderboard this model is not streaming. A live streaming variant of this english ASR model with sub-120ms latency and features such as punctuation, capitalization and custom vocabulary along with other similar models for other languages is available under commercial license from [ABR](https://www.appliedbrainresearch.com/state-space-models).
+The niagara-38m-batch.en model is a [State Space Model](https://huggingface.co/blog/lbourdois/get-on-the-ssm-train) (SSM) with attention that performs automatic speech recognition (ASR), trained and released by [Applied Brain Research](http://www.appliedbrainresearch.com/) (ABR). This model contains \~38m parameters, transcribes speech in English, and was trained on \~50k hours of speech data (competitors use about 200k hours). This is a batch model variant designed for direct comparison with other models on the Open ASR Leaderboard. SSMs are an ideal solution for streaming contexts, but to provide a more direct comparison with other models in the leaderboard this model is not streaming. A live streaming variant of this english ASR model with sub-120ms latency and features such as punctuation, capitalization and custom vocabulary along with other similar models for other languages is available under commercial license from [ABR](https://www.appliedbrainresearch.com/state-space-models).
 
 ## Why “Niagara”?
 
@@ -86,7 +86,7 @@ This model provides transcribed speech as a string for a given audio sample.
 
 The SSM ASR model is trained for English speech recognition and transcribes audio into text. ABR developed the model to demonstrate small, efficient, real-time, accurate speech recognition can be performed with SSMs and run on low cost third party hardware. The model uses 38m parameters. The version posted here is a non-causal model (like most on the leaderboard), to give fair performance comparisons. It is also available as a cascaded model, meaning it produces extremely low latency (\<120ms from first audio to token) causal outputs as well as 1s latency non-causal outputs. These two streams can be merged to have a quick response that updates after 1s with the final result.
 
-Release Date: February 27, 2026
+Release Date: April 15, 2026
 
 ### Model Type
 
@@ -102,7 +102,7 @@ The model was trained on datasets partially listed below. It uses MFCC preproces
 
 ### Datasets
 
-The training datasets contain \~15k hours of English speech, including:
+The training datasets contain \~50k hours of English speech, including:
 
 - LibriSpeech (clean)  
 - VoxPopuli  
@@ -119,43 +119,43 @@ The training datasets contain \~15k hours of English speech, including:
 Our evaluations show that the SSM ASR demonstrates better performance on benchmark datasets compared to other similarly sized or often larger ASR models. The posted model transcribes audio to lower case english with no punctuation. Performance is reported in terms of Word Error Rate (WER%) for the non-causal model.
 
 ```
-Average WER = 8.84 %
+Average WER = 8.91 %
 ```
 
-| Dataset | WER |
-| :---- | :---- |
-| AMI-IHM | 16.81% |
-| Earnings-22 | 10.99% |
-| GigaSpeech | 13.73% |
-| LibriSpeech (clean) | 3.35% |
-| LibriSpeech (other) | 8.80% |
-| SPGISpeech | 2.97% |
-| TED-LIUM | 5.83% |
-| VoxPopuli | 8.24% |
+| Dataset | WER    |
+| :---- |:-------|
+| AMI-IHM | 15.87% |
+| Earnings-22 | 12.09% |
+| GigaSpeech | 12.56% |
+| LibriSpeech (clean) | 3.84%  |
+| LibriSpeech (other) | 9.35%  |
+| SPGISpeech | 3.10%  |
+| TED-LIUM | 5.71%  |
+| VoxPopuli | 8.73%  |
 
 ## Related models
 
 | Model                                                                             | Parameters | Average WER |
-|:----------------------------------------------------------------------------------| :---- | :---- |
-| [abr-ai/niagara-19m-batch.en](https://huggingface.co/abr-ai/niagara-19m-batch.en) | 19M | TBD |
-| [abr-ai/niagara-38m-batch.en](https://huggingface.co/abr-ai/niagara-38m-batch.en) | 38M | TBD |
+|:----------------------------------------------------------------------------------| :---- |:------------|
+| [abr-ai/niagara-19m-batch.en](https://huggingface.co/abr-ai/niagara-19m-batch.en) | 19M | 10.47       |
+| [abr-ai/niagara-38m-batch.en](https://huggingface.co/abr-ai/niagara-38m-batch.en) | 38M | 8.91        |
 
 ## Competitive Comparisons
 
-![][image1]  
+![ASR Leaderboard Data accuracy vs parameter count](plot.png)
 Figure 1: ASR Leaderboard Data accuracy vs parameter count. Transformer-based models marked in grey. SSMs improve the pareto front of transformers.
 
 Models with fewer than 100M parameters, sorted by accuracy.
 
 | Model Name                  | Parameters (M) | Accuracy (%) |
-|:----------------------------| ----- | ----- |
-| abr-ai/niagara-38m-batch.en | 38 | TBD |
-| moonshine-base              | 61 | 90.00 |
-| whisper-base                | 74 | 89.70 |
-| abr-ai/niagara-19m-batch.en | 19 | TBD |
-| moonshine-streaming-tiny    | 34 | 88.00 |
-| moonshine-tiny              | 27 | 87.30 |
-| whisper-tiny                | 39 | 87.30 |
+|:----------------------------| ----- |--------------|
+| abr-ai/niagara-38m-batch.en | 38 | 91.09        |
+| moonshine-base              | 61 | 90.00        |
+| whisper-base                | 74 | 89.70        |
+| abr-ai/niagara-19m-batch.en | 19 | 89.53        |
+| moonshine-streaming-tiny    | 34 | 88.00        |
+| moonshine-tiny              | 27 | 87.30        |
+| whisper-tiny                | 39 | 87.30        |
 
 ## Limitations
 

niagara-38m-batch.en-cpu.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e8c0e30ff39eaa6384c30a72e5fb31995cebefec3cd83bd8bea8290904f9e438
-size 155975431
+oid sha256:ce29f582f057f4fd6d8375b4f093f71f616fe8b7e2b95efcd6480cc87cf8b0a7
+size 155999747

niagara-38m-batch.en-cuda.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ffd05a0776d95a1719183a921c509953ff7bc51f31341f7392cf9665ca02d71d
-size 155978382
+oid sha256:ca7e45f5acac5bf0ae9b9a4737999ab037a474444bbcdb70368deb038ba80dc6
+size 156002698

sentencepiece_1024.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c8be70290b623a76004f077f56c90d14f9eb098b813b3997a6a460dcaa2e5ce8
-size 254559
+oid sha256:80bdd5b93f5688ea8e5a659d165f087877c5868497a0b818dfe11a919770010b
+size 254572