// poolside Laguna · Text LLM
Laguna-S-2.1: VRAM & local deployment
118B total · 8B active (MoE) parameters. About 79.7 GB at 4-bit and 261 GB at BF16 with one 8K-token request.
released 2026-07-13licence: Commercial OKmemory figures are estimates
01Specs
| Publisher | poolside · poolside |
|---|---|
| Family / series | poolside Laguna · Laguna-S-2.1 |
| Type | Text LLM (text-generation) |
| Parameters | 118B total · 8B active (MoE) Source: official model card“118B total, ~8B activated per token” |
| Context length | 1M tokens (1,048,576)max_position_embeddings in config.json |
| Architecture | 48 layers (12 full-attention) · 36 sliding-window layers (512) · GQA · MoE |
| Released weights | BF16 · 235 GBdownload |
| Licence | OpenMDW-1.1 · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed. |
| Released | 2026-07-13 (Hugging Face repo created) |
02How much VRAM it needs (estimate)
Total = weights + 10% + 0.5GB overhead + KV cache for one request (FP16 KV). More concurrent users multiply the KV column.
| Precision | Weights | Total @ 8K | Total @ 32K | Total @ 128K | Smallest GPU | Mac |
|---|---|---|---|---|---|---|
| BF16 / FP16 (original) | 236 GB | 261 GB | 262 GB | 267 GB | 4 × RTX PRO 6000 (384GB) | Mac 512GB |
| FP8 / INT8 / Q8_0 | 125 GB | 139 GB | 140 GB | 145 GB | 4 × RTX PRO 6000 (384GB) | Mac 256GB |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 71.5 GB | 79.7 GB | 80.9 GB | 85.7 GB | RTX PRO 6000 96GB | Mac 128GB |
KV cache per request: 8K 0.5 GB · 32K 1.7 GB · 128K 6.5 GB
Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →
03How to run it
Only runtimes the publisher or the runtime’s official library names for this model.
- vLLM — mentioned in the official model card
- SGLang — mentioned in the official model card
- llama.cpp — mentioned in the official model card
- MLX — mentioned in the official model card
- Transformers — mentioned in the official model card
- TensorRT-LLM — mentioned in the official model card
- Ollama — in the official Ollama library · laguna-s-2.1
ollama run laguna-s-2.104Official repositories & quantised builds
9 official repo(s) from poolside.
- Laguna-S-2.1-GGUF GGUF
- Laguna-S-2.1 main
- Laguna-S-2.1-NVFP4 NVFP4
- Laguna-S-2.1-DFlash-NVFP4 NVFP4
- Laguna-S-2.1-FP8 FP8
- Laguna-S-2.1-INT4 INT4
- Laguna-S-2.1-NVFP4-mlx NVFP4/MLX
- Laguna-S-2.1-DFlash-FP8 FP8
- Laguna-S-2.1-DFlash-INT4 INT4
05Other poolside Laguna models
All 4 →| Model | Params | 4-bit total | Released |
|---|---|---|---|
| Laguna-XS-2.1 | 33B / 3B | 22.9 GB | 2026-06-20 |
| Laguna-M.1 | 225B / 23B | 153 GB | 2026-06-15 |
| Laguna-XS.2 | 33B / 3B | 22.9 GB | 2026-04-23 |
06Sources
Data checked 2026-08-19. Parameters, context and licence are copied from the official repository; memory figures are our estimates.