GigaAM v3 RNN-T needs about 0.7 GB of memory at Q4_K_M (a 0.2 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 6 quants, from 0.2 GB (Q4_K_M) to 0.9 GB (F32).
GigaAM v3 RNN‑T: 220‑M parameter Russian speech recognizer
GigaAM v3 RNN‑T is a Conformer‑based ASR model with 220‑240 M parameters. It is fine‑tuned with an RNN‑T decoder for greedy decoding. The model outputs lowercased Russian text without punctuation.
Good at
- Russian speech‑to‑text
- Low‑latency greedy decoding
- Small model size
Limits
- No vision support
- Limited to 8 k context
- MIT license only for text use
Best for
- Transcribing short Russian audio clips
- Batching small audio files
- Offline speech recognition on modest hardware
Who it’s for. Developers needing a lightweight Russian ASR model for offline use. The model is suitable for embedded systems or edge devices that require quick transcription without external language models.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 0.2 GB | 0.7 GB | Fits | Get file → | |
| Q5_K_M | 0.2 GB | 0.7 GB | Fits | Get file → | |
| Q6_K | 0.2 GB | 0.8 GB | Fits | Get file → | |
| Q8_0 | 0.3 GB | 0.8 GB | Fits | Get file → | |
| F16 | 0.5 GB | 1.0 GB | Fits | Get file → | |
| F32 | 0.9 GB | 1.5 GB | Fits | Get file → |
For 8 GB VRAM, Q4_K_M (0.18 GB) or Q5_K_M (0.21 GB) are recommended, both using about 0.7 GB memory at 8 k context. If 4 GB VRAM is available, Q4_K_M is the best choice. Higher precision quants increase memory usage and are unnecessary for most use cases.
What GPU do you need for GigaAM v3 RNN-T?
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 F32 (0.9 GB file, about 1.5 GB in use).
Which GigaAM v3 RNN-T quant should you choose?
Q4_K_M is the usual starting point for GigaAM v3 RNN-T: it balances file size and quality well. Here it is a 0.2 GB file.
The Q8_0 file is 0.3 GB, about 1.5 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about GigaAM v3 RNN-T
How much VRAM does GigaAM v3 RNN-T need?
About 0.7 GB at Q4_K_M with an 8K context window. That is the 0.2 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 GigaAM v3 RNN-T on an 8 GB GPU?
Yes, with F32: the 0.9 GB file needs about 1.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of GigaAM v3 RNN-T GGUF?
It depends on the quant: Q4_K_M is 0.2 GB, Q5_K_M is 0.2 GB, Q6_K is 0.2 GB, Q8_0 is 0.3 GB, F16 is 0.5 GB, F32 is 0.9 GB.
Where can I download GigaAM v3 RNN-T GGUF?
The files are on Hugging Face in the handy-computer/gigaam-v3-rnnt-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 GigaAM v3 RNN-T use?
Hugging Face lists the license as mit. 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 GigaAM v3 RNN‑T on 4 GB VRAM?
Yes, the Q4_K_M quant uses about 0.7 GB memory at 8 k context, fitting comfortably in 4 GB VRAM.
Does this model support punctuation?
No, the RNN‑T variant outputs lowercased Russian text without punctuation.
What is the context length limit?
The model is limited to an 8 k token context length.