cstr/parakeet-ctc-1.1b-GGUF

Parakeet CTC 1.1B — GGUF (ggml-quantised)

GGUF / ggml conversions of nvidia/parakeet-ctc-1.1b for use with the crispasr CLI from CrispStrobe/CrispASR.

Parakeet CTC 1.1B is NVIDIA's 1.1 B-parameter English ASR model — the deeper sibling of parakeet-ctc-0.6b:

• English-only, lowercase output
• FastConformer encoder — 42 layers (vs 24 in 0.6b), d_model=1024, 8 heads + single CTC head; trained on 64K hours of English (40K private + 24K public)
• CC-BY-4.0 licence
• WERs (slightly better than 0.6b across the board): LibriSpeech-clean 1.83 %, LibriSpeech-other 3.54 %, TEDLIUM-v3 4.20 %, GigaSpeech 10.27 %, Common Voice 6.53 %

This repo provides four quantisations, all converted from the upstream .nemo checkpoint via models/convert-stt-fastconformer-ctc-to-gguf.py and quantised with crispasr-quantize.

Files

File	Size	Notes
parakeet-ctc-1.1b.gguf	~2.13 GB	F16, full precision
parakeet-ctc-1.1b-q8_0.gguf	~1.26 GB	Q8_0, near-lossless
parakeet-ctc-1.1b-q5_0.gguf	~910 MB	Q5_0
parakeet-ctc-1.1b-q4_k.gguf	~795 MB	Q4_K — recommended default

All quantisations produce the same JFK 11 s transcript.

Quick start

# 1. Build the runtime
git clone https://github.com/CrispStrobe/CrispASR
cd CrispASR
cmake -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j$(nproc) --target crispasr-cli

# 2. Run — the CLI auto-downloads Q4_K from this repo by friendly name:
./build/bin/crispasr -m parakeet-ctc-1.1b -f your-audio.wav

# Or pre-download a specific quant via huggingface_hub and point to it:
python -c "from huggingface_hub import hf_hub_download; print(hf_hub_download('cstr/parakeet-ctc-1.1b-GGUF', 'parakeet-ctc-1.1b-q8_0.gguf'))"
./build/bin/crispasr -m parakeet-ctc-1.1b-q8_0.gguf -f your-audio.wav

The crispasr CLI auto-detects the backend from filename — parakeet-ctc-*.gguf routes to fastconformer-ctc because the architecture is identical to the stt_en_fastconformer_ctc_* family. Registry key parakeet-ctc-1.1b triggers Q4_K auto-download.

Model architecture

Component	Details
Encoder	42-layer FastConformer, d=1024, 8 heads, head_dim=128, FFN=4096, conv kernel=9, attention biases ON
Subsampling	dw_striding stack, 8× temporal (50 → 12.5 fps)
CTC head	Conv1d(1024 → 1025), k=1; vocab 1024 SentencePiece + 1 blank
Audio	16 kHz mono, 80 mel bins, n_fft=512, hop=160, win=400
Parameters	~1.1 B

Identical to the 0.6b variant in every dimension except encoder depth (42 vs 24 layers). The crispasr fastconformer-ctc backend reads n_layers from the GGUF metadata and resizes per-layer state at load time, so no code changes are needed for the deeper model.

Output convention

Lowercase, un-punctuated English. For cased + punctuated, see cstr/parakeet-tdt-0.6b-v3-GGUF (multilingual TDT, slightly higher WER but cased+punct out-of-the-box) or post-process via crispasr's --punc-model.

Attribution

• Original model: nvidia/parakeet-ctc-1.1b (CC-BY-4.0). NVIDIA NeMo team.
• GGUF conversion + ggml runtime: CrispStrobe/CrispASR — FastConformer-CTC backend, see src/canary_ctc.cpp.

Related

• C++ runtime: CrispStrobe/CrispASR
• Smaller CTC variant: cstr/parakeet-ctc-0.6b-GGUF
• TDT counterpart (multilingual, cased+punct): cstr/parakeet-tdt-0.6b-v3-GGUF

License

CC-BY-4.0, inherited from the base model. Use of these GGUF files must comply with the CC-BY-4.0 license including attribution.