Kolibri 1 needs about 30.8 GB of memory at IQ3_XXS (a 29.5 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 3 quants, from 21.5 GB (IQ2_XS) to 39.0 GB (IQ4_XS). Longer context needs more: at 32K tokens the KV cache alone adds about 3.4 GB.
Kolibri 1 – 74.4B parameter MoE model for German and English
Kolibri 1 is a mixture-of-experts model with 74.4B parameters, optimized for long‑context inference. It supports explicit reasoning mode and tool calling. The model is available in three GGUF quantizations.
Good at
- Long‑context reasoning
- Tool calling
- German and English text processing
Limits
- Requires patched llama.cpp for kolibri1 architecture
- Full model must fit in memory
- No vision support
Best for
- Conversational assistants
- Document‑processing workflows
- Question‑answering over internal data
Who it’s for. Developers building conversational assistants or agentic workflows that need German and English language support. The model is intended for use in systems where a human reviews output before action.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_XS | 21.5 GB | 22.8 GB | Too big | Get file → | |
| IQ3_XXS | 29.5 GB | 30.8 GB | Too big | Get file → | |
| IQ4_XS | 39.0 GB | 40.3 GB | Too big | Get file → |
For 24 GB VRAM use IQ2_XS (21.47 GB file, 22.8 GB memory at 8 k context). For 32 GB VRAM use IQ3_XXS (29.47 GB file, 30.8 GB memory). For 48 GB VRAM use IQ4_XS (38.98 GB file, 40.3 GB memory). Estimates are approximate.
What GPU do you need for Kolibri 1?
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: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: the best quant that fits is IQ2_XS (21.5 GB file, about 22.8 GB in use).
- 32 GB VRAM: the best quant that fits is IQ3_XXS (29.5 GB file, about 30.8 GB in use).
- 48 GB VRAM: the best quant that fits is IQ4_XS (39.0 GB file, about 40.3 GB in use).
Which Kolibri 1 quant should you choose?
IQ3_XXS (29.5 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 IQ2_XS at 21.5 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 Kolibri 1
How much VRAM does Kolibri 1 need?
About 30.8 GB at IQ3_XXS with an 8K context window. That is the 29.5 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 Kolibri 1 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 22.8 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Kolibri 1 GGUF?
It depends on the quant: IQ2_XS is 21.5 GB, IQ3_XXS is 29.5 GB, IQ4_XS is 39.0 GB.
What context length does Kolibri 1 support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.8 GB at 8K and 3.4 GB at 32K.
Where can I download Kolibri 1 GGUF?
The files are on Hugging Face in the webmp3/Sakura-MicroQuality-Kolibri-1-365E-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 Kolibri 1 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 Kolibri 1 on 8 GB VRAM?
No, the smallest quant requires at least 22.8 GB memory at 8 k context.
Does Kolibri 1 support vision tasks?
No, the model is text‑only.
Do I need a patched llama.cpp to run this model?
Yes, the kolibri1 architecture requires a patched llama.cpp as described in the repository notes.