Laguna S 2.1, GGUF
Laguna S 2.1 is a Mixture-of-Experts language model designed for agentic coding, reasoning, and long-horizon workflows. The model has grouped-query attention and per-head gating. It is optimized for large-scale coding, tool use, and long-context reasoning.
This repository holds Vultr's GGUF quantizations of it. They are hybrid quantizations: rather than applying one precision to the whole network, each layer's precision is chosen from a measured importance score, so the bits go where they change the output most.
Model Overview
| Attribute | Value |
|---|---|
| Base model | poolside/Laguna-S-2.1 |
| Architecture | laguna, Mixture-of-Experts |
| Parameters | 118B total |
| Active parameters | approximately 8B per token |
| Layers | 48 |
| Hidden size | 3,072 |
| Attention heads | 48 |
| Experts | 256 routed, 10 selected per token, plus a shared expert |
| Context length | up to 1M tokens |
| Vocabulary | 100,352 |
| License | OpenMDW-1.1 |
| Attribute | Value |
|---|---|
| Base model | poolside/Laguna-S-2.1 |
| Architecture | laguna, Mixture-of-Experts |
| Parameters | 118B total |
| Active parameters | approximately 8B per token |
| Layers | 48 |
| Hidden size | 3,072 |
| Attention heads | 48 |
| Experts | 256 routed, 10 selected per token, plus a shared expert |
| Context length | up to 1M tokens |
| Vocabulary | 100,352 |
| License | OpenMDW-1.1 |
Quantizations
| File | Type | Size | KLD vs bf16 |
|---|---|---|---|
| Laguna-S-2.1-Q5_K_M.gguf | Q5_K_M | 91.76 GB | 0.0254 |
| Laguna-S-2.1-Q4_K_M.gguf | Q4_K_M | 78.26 GB | 0.0588 |
| Laguna-S-2.1-IQ3_M.gguf | IQ3_M | 57.16 GB | 0.1288 |
| Laguna-S-2.1-IQ3_S.gguf | IQ3_S | 56.38 GB | 0.1503 |
| File | Type | Size | KLD vs bf16 |
|---|---|---|---|
| Laguna-S-2.1-Q5_K_M.gguf | Q5_K_M | 91.76 GB | 0.0254 |
| Laguna-S-2.1-Q4_K_M.gguf | Q4_K_M | 78.26 GB | 0.0588 |
| Laguna-S-2.1-IQ3_M.gguf | IQ3_M | 57.16 GB | 0.1288 |
| Laguna-S-2.1-IQ3_S.gguf | IQ3_S | 56.38 GB | 0.1503 |
KL divergence measures how far a quantized model's next-token probabilities drift from those of the original bf16 model, averaged over every token of the evaluation text. The lower the value, the closer the quantized model is to the original; zero would mean identical output.
Evaluation
Every variant in this repository was evaluated against the original bf16 model for perplexity, KL divergence, top-token agreement, HellaSwag and WinoGrande, under identical conditions. The full results are maintained on the model's page in the Vultr Model Library:
https://docs.vultr.com/models/poolside-laguna-s-21/laguna-s-2.1?tab=quantized
That page is the reference for these files. The summary above is an excerpt from it.
Run
llama-server -hf vultr/Laguna-S-2.1-GGUF:Q4_K_M
The model's chat template is embedded in every file, so no extra prompt formatting is needed.
License
These files are derived from poolside/Laguna-S-2.1 and are distributed under the same license, OpenMDW-1.1. The model is the work of Poolside; Vultr produced the quantizations.