Vikhr-Gemma-2B-instruct needs about 3.1 GB of memory at Q4_K_M (a 1.7 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 33 quants, from 0.8 GB (IQ1_S) to 10.5 GB (F32).
Vikhr-Gemma-2B-instruct – a compact Russian‑focused instruct model
Vikhr-Gemma-2B-instruct is a Gemma 2 2B‑based model trained on the GrandMaster‑PRO‑MAX Russian dataset. It is optimized for instruction following and concise responses. The model supports a 8192‑token context window and is released under Apache‑2.0.
Good at
- Instruction following
- Russian language generation
- Compact size
Limits
- No vision support
- Limited to 8192‑token context
- Requires 2.2‑3.5 GB VRAM for 8k context depending on quant
Best for
- Chatbots needing Russian instructions
- Quick prototyping of Russian LLM apps
- Low‑memory inference on GPUs
Who it’s for. Developers building Russian‑language assistants or educational tools. Researchers needing a small, instruction‑tuned Gemma variant for experimentation. Users with modest GPU memory who still want decent performance on Russian text.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_S | 0.8 GB | 2.2 GB | Fits | Get file → | |
| IQ1_M | 0.9 GB | 2.2 GB | Fits | Get file → | |
| IQ2_XXS | 0.9 GB | 2.3 GB | Fits | Get file → | |
| IQ2_XS | 1.0 GB | 2.4 GB | Fits | Get file → | |
| IQ2_S | 1.0 GB | 2.4 GB | Fits | Get file → | |
| IQ2_M | 1.1 GB | 2.5 GB | Fits | Get file → | |
| Q2_K_S | 1.2 GB | 2.5 GB | Fits | Get file → | |
| IQ3_XXS | 1.2 GB | 2.6 GB | Fits | Get file → | |
| Q2_K | 1.2 GB | 2.6 GB | Fits | Get file → | |
| IQ3_XS | 1.3 GB | 2.7 GB | Fits | Get file → | |
| IQ3_S | 1.4 GB | 2.7 GB | Fits | Get file → | |
| Q3_K_S | 1.4 GB | 2.7 GB | Fits | Get file → | |
| IQ3_M | 1.4 GB | 2.8 GB | Fits | Get file → | |
| Q3_K | 1.5 GB | 2.8 GB | Fits | Get file → | |
| Q3_K_M | 1.5 GB | 2.8 GB | Fits | Get file → | |
| Q3_K_L | 1.6 GB | 2.9 GB | Fits | Get file → | |
| IQ4_XS | 1.6 GB | 2.9 GB | Fits | Get file → | |
| Q4_0 | 1.6 GB | 3.0 GB | Fits | Get file → | |
| IQ4_NL | 1.6 GB | 3.0 GB | Fits | Get file → | |
| Q4_K_S | 1.6 GB | 3.0 GB | Fits | Get file → | |
| Q4_K | 1.7 GB | 3.1 GB | Fits | Get file → | |
| Q4_K_M | 1.7 GB | 3.1 GB | Fits | Get file → | |
| Q4_1 | 1.8 GB | 3.1 GB | Fits | Get file → | |
| Q5_0 | 1.9 GB | 3.3 GB | Fits | Get file → | |
| Q5_K_S | 1.9 GB | 3.3 GB | Fits | Get file → | |
| Q5_K | 1.9 GB | 3.3 GB | Fits | Get file → | |
| Q5_K_M | 1.9 GB | 3.3 GB | Fits | Get file → | |
| Q5_1 | 2.0 GB | 3.4 GB | Fits | Get file → | |
| Q6_K | 2.2 GB | 3.5 GB | Fits | Get file → | |
| Q8_0 | 2.8 GB | 4.2 GB | Fits | Get file → | |
| BF16 | 5.2 GB | 6.6 GB | Fits | Get file → | |
| F16 | 5.2 GB | 6.6 GB | Fits | Get file → | |
| F32 | 10.5 GB | 11.8 GB | Too big | Get file → |
For 8 GB VRAM, use IQ1_S or IQ1_M (file size 0.83 GB, memory 2.2 GB). If 12 GB VRAM is available, Q4_K or Q4_K_M (file size 1.71 GB, memory 3.1 GB) offers better quality. For 16 GB VRAM, Q5_K or Q5_K_M (file size 1.92 GB, memory 3.3 GB) provides a balance of quality and memory usage.
What GPU do you need for Vikhr-Gemma-2B-instruct?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) 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 Q6_K (2.2 GB file, about 3.5 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (2.8 GB file, about 4.2 GB in use).
- 8 GB VRAM: the best quant that fits is BF16 (5.2 GB file, about 6.6 GB in use).
- 12 GB VRAM: the best quant that fits is F32 (10.5 GB file, about 11.8 GB in use).
Which Vikhr-Gemma-2B-instruct quant should you choose?
Q4_K_M is the usual starting point for Vikhr-Gemma-2B-instruct: it balances file size and quality well. Here it is a 1.7 GB file.
The Q8_0 file is 2.8 GB, about 1.6 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.
If memory is tight, the smallest file is IQ1_S at 0.8 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Vikhr-Gemma-2B-instruct
How much VRAM does Vikhr-Gemma-2B-instruct need?
About 3.1 GB at Q4_K_M with an 8K context window. That is the 1.7 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 Vikhr-Gemma-2B-instruct on an 8 GB GPU?
Yes, with BF16: the 5.2 GB file needs about 6.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Vikhr-Gemma-2B-instruct GGUF?
It depends on the quant: IQ1_S is 0.8 GB, IQ1_M is 0.9 GB, IQ2_XXS is 0.9 GB, IQ2_XS is 1.0 GB, IQ2_S is 1.0 GB, IQ2_M is 1.1 GB and 27 more.
What context length does Vikhr-Gemma-2B-instruct support?
The GGUF header lists a trained context length of 8,192 tokens (8K). The KV cache grows with the context you use: about 0.9 GB at 8K and 0.9 GB at 8K.
Where can I download Vikhr-Gemma-2B-instruct GGUF?
The files are on Hugging Face in the Vikhrmodels/Vikhr-Gemma-2B-instruct-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 Vikhr-Gemma-2B-instruct use?
Hugging Face lists the license as apache-2.0. 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 Vikhr-Gemma-2B-instruct on 8 GB VRAM?
Yes, with IQ1_S or IQ1_M quantization the model uses about 2.2 GB of VRAM at 8k context.
Does the model support English?
The model is specialized for Russian; English performance is not guaranteed.
Is the model free to use?
Yes, it is released under the Apache‑2.0 license.