addis scribe streaming needs about 0.9 GB of memory at Q3_K (a 0.3 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 22 quants, from 0.0 GB (Q3_K) to 0.7 GB (Q8_0).
Addis Scribe Streaming – real‑time Amharic ASR on NVIDIA Nemotron 3.5
Addis Scribe Streaming is a 0.0993B parameter model built on NVIDIA Nemotron 3.5 ASR Streaming. It delivers real‑time Amharic speech recognition with 0.32‑second steps and final output 30 ms after speech stops. The model supports 40 language‑locales via language‑ID prompt conditioning.
Good at
- Real‑time Amharic transcription
- Low‑latency streaming
- Multilingual support via language‑ID conditioning
Limits
- No punctuation support
- WER about 3 points higher than offline models on natural speech
- Requires NVIDIA NeMo 3.0 for quickstart
Best for
- Live voice assistants in Amharic
- Batch transcription of Amharic audio
- Multilingual ASR pipelines
Who it’s for. Developers building voice agents or transcription services that need Amharic support. The model is ready for commercial use and can be integrated into streaming applications.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q3_K | 0.0 GB | 0.6 GB | Fits | Get file → | |
| IQ4_NL | 0.1 GB | 0.6 GB | Fits | Get file → | |
| Q4_0 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| Q4_K | 0.1 GB | 0.6 GB | Fits | Get file → | |
| Q5_0 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| Q5_K | 0.1 GB | 0.6 GB | Fits | Get file → | |
| Q6_K | 0.1 GB | 0.6 GB | Fits | Get file → | |
| Q8_0 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| Q8_0 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| — | 0.3 GB | 0.9 GB | Fits | Get file → | |
| Q3_K | 0.3 GB | 0.9 GB | Fits | Get file → | |
| — | 0.4 GB | 0.9 GB | Fits | Get file → | |
| IQ4_NL | 0.4 GB | 1.0 GB | Fits | Get file → | |
| Q4_0 | 0.4 GB | 1.0 GB | Fits | Get file → | |
| — | 0.4 GB | 1.0 GB | Fits | Get file → | |
| Q4_K | 0.4 GB | 1.0 GB | Fits | Get file → | |
| Q5_0 | 0.5 GB | 1.0 GB | Fits | Get file → | |
| Q5_K | 0.5 GB | 1.1 GB | Fits | Get file → | |
| Q5_K | 0.5 GB | 1.1 GB | Fits | Get file → | |
| Q6_K | 0.6 GB | 1.1 GB | Fits | Get file → | |
| Q8_0 | 0.7 GB | 1.3 GB | Fits | Get file → | |
| Q8_0 | 0.7 GB | 1.3 GB | Fits | Get file → |
For 8 GB VRAM, use Q3_K (0.04 GB file, 0.6 GB memory at 8k context) or IQ4_NL (0.06 GB file, 0.6 GB memory). For 12 GB VRAM, Q5_K (0.5 GB file, 1.1 GB memory) or Q6_K (0.58 GB file, 1.1 GB memory) provide a balance of size and performance.
What GPU do you need for addis scribe streaming?
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 Q8_0 (0.7 GB file, about 1.3 GB in use).
Which addis scribe streaming quant should you choose?
Q3_K (0.3 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
If memory is tight, the smallest file is Q3_K at 0.0 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about addis scribe streaming
How much VRAM does addis scribe streaming need?
About 0.9 GB at Q3_K with an 8K context window. That is the 0.3 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 addis scribe streaming on an 8 GB GPU?
Yes, with Q8_0: the 0.7 GB file needs about 1.3 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of addis scribe streaming GGUF?
It depends on the quant: Q3_K is 0.0 GB, IQ4_NL is 0.1 GB, Q4_0 is 0.1 GB, Q4_K is 0.1 GB, Q5_0 is 0.1 GB, Q5_K is 0.1 GB and 16 more.
Where can I download addis scribe streaming GGUF?
The files are on Hugging Face in the addisai/addis-scribe-streaming repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does addis scribe streaming use?
Hugging Face lists the license as other. 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 Addis Scribe Streaming on 8 GB VRAM?
Yes, the Q3_K and IQ4_NL quantizations use 0.6 GB memory at 8k context, fitting within 8 GB VRAM.
Does the model support punctuation?
No, the training text had punctuation removed, so the model does not output punctuation.
What is the latency of the streaming inference?
Text appears in 0.32‑second steps and final output is about 30 ms after speech stops.