Laguna Xs21 Hf needs about 22.1 GB of memory at Q4_K_M (a 20.3 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 5 quants, from 16.1 GB (Q3_K_M) to 35.6 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 5.4 GB.
Laguna Xs21 Hf – 33B Mixture‑of‑Experts LLM for long‑horizon coding
Laguna Xs21 Hf is a 33.4B parameter Mixture‑of‑Experts model with 3B activated per token. It supports a 262,144 token context window and native reasoning with interleaved thinking. The model is fully offline and available in multiple low‑bit GGUF quantizations.
Good at
- Long‑horizon coding
- Agentic reasoning
- Terminal‑style tasks
Limits
- Requires 36 GB RAM for full 33B model
- No vision or multimodal support
- OpenMDW‑1.1 license allows commercial use
Best for
- Local code generation
- AI agent workflows
- Large‑context document analysis
Who it’s for. Developers and researchers who need a powerful local LLM for coding and reasoning tasks. The model runs on a Mac with 36 GB of RAM and is available via Ollama, llama.cpp, or Atomic Chat.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q3_K_M | 16.1 GB | 17.9 GB | Too big | Get file → | |
| Q4_K_M | 20.3 GB | 22.1 GB | Too big | Get file → | |
| Q5_K_M | 23.8 GB | 25.6 GB | Too big | Get file → | |
| Q6_K | 27.5 GB | 29.3 GB | Too big | Get file → | |
| Q8_0 | 35.6 GB | 37.4 GB | Too big | Get file → |
For 8 GB VRAM use Q3_K_M (16.06 GB file, ~17.9 GB memory). For 12 GB VRAM use Q4_K_M (20.27 GB file, ~22.1 GB memory). For 16 GB VRAM use Q5_K_M (23.78 GB file, ~25.6 GB memory).
What GPU do you need for Laguna Xs21 Hf?
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 Q4_K_M (20.3 GB file, about 22.1 GB in use).
- 32 GB VRAM: the best quant that fits is Q6_K (27.5 GB file, about 29.3 GB in use).
- 48 GB VRAM: the best quant that fits is Q8_0 (35.6 GB file, about 37.4 GB in use).
Which Laguna Xs21 Hf quant should you choose?
Q4_K_M is the usual starting point for Laguna Xs21 Hf: it balances file size and quality well. Here it is a 20.3 GB file.
The Q8_0 file is 35.6 GB, about 1.8 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 Q3_K_M at 16.1 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 Laguna Xs21 Hf
How much VRAM does Laguna Xs21 Hf need?
About 22.1 GB at Q4_K_M with an 8K context window. That is the 20.3 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 Laguna Xs21 Hf on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 17.9 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Laguna Xs21 Hf GGUF?
It depends on the quant: Q3_K_M is 16.1 GB, Q4_K_M is 20.3 GB, Q5_K_M is 23.8 GB, Q6_K is 27.5 GB, Q8_0 is 35.6 GB.
What context length does Laguna Xs21 Hf support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.3 GB at 8K and 5.4 GB at 32K.
Where can I download Laguna Xs21 Hf GGUF?
The files are on Hugging Face in the AtomicChat/Laguna-XS-2.1-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 Laguna Xs21 Hf use?
Hugging Face lists the license as openmdw-1.1. 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 Laguna Xs21 Hf on 8 GB VRAM?
Use the Q3_K_M quant, which requires about 17.9 GB of memory at 8 k context. 8 GB VRAM is insufficient for this model.
Does the model support image input?
No, Laguna Xs21 Hf is text‑to‑text only.
What license applies to this model?
It is released under OpenMDW‑1.1, fully permissive for commercial and non‑commercial use.