cstr/ark-asr-3b-GGUF
automatic-speech-recognition model by cstr
From the publisher
Model card & documentation
Source previewRead the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.
GGUF conversions of AutoArk-AI/ARK-ASR-3B, a 19-language automatic speech recognition model, for use with CrispASR (a ggml/llama.cpp-style C++ engine). ARK-ASR-3B = Whisper-large-v3 encoder with partial rotary position embeddings → MLP adapter → Qwen2.5-3B decoder with audio-token injection. A single self-contained GGUF holds the encoder, adapter, and language model. positional embeddings are replaced by partial interleaved RoPE (rotates the first 32 of 64 head dims, θ=10000, on Q and K; kproj has no bias; the encoder's own final LayerNorm is dropped). Because positions are rotary, the encoder handles arbitrary-length audio in one pass — there is no 30 s cap. → Linear 4096→2048. RMSNorm, θ=1e6, tied embeddings). The placeholder embeddings are overwritten by the adapter's audio features, then the transcript is decoded. Mel features use the stock WhisperFeatureExtractor recipe (128 bins, nfft 400, hop 160). All three transcribe the JFK sample verbatim. Q80 is the best size/quality trade-off. -m auto --backend ark-asr…
Read the full model card ↗ · Preview checked 2026-11-20T07:15:14.246Z
Inside the original model card — Document outline
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Documentation belongs to its respective authors. Reported project/model license: other. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
cstr/ark-asr-3b-GGUF is a automatic-speech-recognition model repository published by cstr on Hugging Face. The source reports the gguf library.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- automatic-speech-recognition
- Library
- gguf
- Recent downloads (30 days)
- 197,719
- Cumulative likes
- 2
- Hugging Face trending score
- 0
- Reported safetensors parameters
- Not reported
- Architecture
- Not reported
- License
- other
- Access
- Not gated by Hugging Face
- Created
- 2026-06-29T06:25:58.000Z
- Last modified
- 2026-07-03T14:22:42.000Z
Compatibility and lineage
Reported languages: zh, en, de, ja, fr, ko, es, pl, it, ro, hu, cs, nl, fi, hr, sk, sl, et, lt
Reported base models: AutoArk-AI/ARK-ASR-3B
Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.
Use and evaluate
Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.
No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.
Source and freshness
Source: Hugging Face Hub. Metadata observed 2026-11-20T07:04:13.055Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.
Related models
jonatasgrosman/wav2vec2-large-xlsr-53-japanese