| --- |
| library_name: mlx |
| license: mit |
| language: |
| - ru |
| - en |
| tags: |
| - automatic-speech-recognition |
| - mlx |
| - apple-silicon |
| - russian |
| - gigaam |
| - conformer |
| - ctc |
| base_model: ai-sage/GigaAM-v3 |
| pipeline_tag: automatic-speech-recognition |
| model-index: |
| - name: GigaAM-v3-e2e-ctc-mlx |
| results: |
| - task: |
| type: automatic-speech-recognition |
| metrics: |
| - name: RTF (M2 Max) |
| type: rtf |
| value: 0.006 |
| --- |
| |
| # GigaAM v3 e2e CTC — MLX |
|
|
| MLX port of [GigaAM-v3](https://github.com/salute-developers/GigaAM) for fast Russian speech recognition on Apple Silicon. **180x realtime** on M2 Max. |
|
|
| ## Usage |
|
|
| ```bash |
| pip install gigaam-mlx |
| ``` |
|
|
| ```python |
| from gigaam_mlx import load_model, transcribe |
| |
| model, tokenizer = load_model() # downloads weights automatically |
| text = transcribe(model, tokenizer, "recording.wav") |
| print(text) |
| ``` |
|
|
| Or via CLI: |
|
|
| ```bash |
| gigaam-mlx recording.wav |
| ``` |
|
|
| ## Performance |
|
|
| MacBook Pro M2 Max, 20-second chunk: |
|
|
| | Backend | Time | Realtime | |
| |---|---|---| |
| | **MLX CTC (this)** | **0.11s** | **180x** | |
| | PyTorch MPS RNNT | 0.76s | 26x | |
| | ONNX CPU CTC | 1.66s | 12x | |
|
|
| ## Model |
|
|
| - **Architecture:** Conformer (16 layers, 768d, 16 heads, RoPE) + CTC |
| - **Parameters:** 220M |
| - **Vocabulary:** 257 tokens (SentencePiece) |
| - **Features:** Punctuation, text normalization, Russian + English code-switching |
|
|
| ## Links |
|
|
| - **Code:** [github.com/aystream/gigaam-mlx](https://github.com/aystream/gigaam-mlx) |
| - **Original:** [salute-developers/GigaAM](https://github.com/salute-developers/GigaAM) ([paper](https://arxiv.org/abs/2506.01192)) |
| - **License:** MIT |
|
|
