Kyutai-STT needs about 1.8 GB of memory at Q8_0 (a 1.2 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 0.7 GB (Q4_K) to 2.1 GB.
Kyutai-STT 1B: Streaming speech-to-text for English and French
Kyutai-STT 1B is a 1.05B parameter transformer model for streaming speech-to-text. It uses a Mimi audio codec encoder and a 16‑layer causal transformer LM. The model outputs text with capitalization and punctuation and supports real‑time transcription.
Good at
- Real‑time speech transcription
- English and French languages
- Low‑latency output
Limits
- No vision support
- Requires 8k context for memory estimates
- CC‑BY‑4.0 license
Best for
- Live captioning
- Transcribing podcasts
- Real‑time meeting notes
Who it’s for. Developers building real‑time transcription services or applications that need low‑latency speech-to-text in English or French. The model is suitable for integration with tools that support GGUF and llama.cpp.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K | 0.7 GB | 1.2 GB | Fits | Get file → | |
| Q8_0 | 1.2 GB | 1.8 GB | Fits | Get file → | |
| — | 2.1 GB | 2.9 GB | Fits | Get file → |
For 8 GB VRAM use Q4_K (0.67 GB, ~1.2 GB RAM). For 12 GB VRAM use Q8_0 (1.17 GB, ~1.8 GB RAM). The unknown quant is larger (2.12 GB, ~2.9 GB RAM).
What GPU do you need for Kyutai-STT?
These figures add the file size, an fp16 KV cache at an 8K context window and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is the main file (2.1 GB file, about 2.9 GB in use).
Which Kyutai-STT quant should you choose?
Q8_0 (1.2 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
If memory is tight, the smallest file is Q4_K at 0.7 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Kyutai-STT
How much VRAM does Kyutai-STT need?
About 1.8 GB at Q8_0 with an 8K context window. That is the 1.2 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Kyutai-STT on an 8 GB GPU?
Yes, with the smaller files: the 2.1 GB file needs about 2.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Kyutai-STT GGUF?
It depends on the quant: Q4_K is 0.7 GB, Q8_0 is 1.2 GB, file is 2.1 GB.
Where can I download Kyutai-STT GGUF?
The files are on Hugging Face in the cstr/kyutai-stt-1b-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Kyutai-STT use?
Hugging Face lists the license as cc-by-4.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Kyutai-STT on 8 GB VRAM?
Yes, the Q4_K quantized model is 0.67 GB and requires about 1.2 GB of RAM at 8k context.
Does Kyutai-STT support French?
Yes, the 1B variant supports English and French transcription.