// poolside Laguna · Text LLM
Laguna-XS.2: VRAM & local deployment
33B total · 3B active (MoE) parameters. About 22.9 GB at 4-bit and 73.5 GB at BF16 with one 8K-token request.
released 2026-04-23licence: Commercial OKmemory figures are estimates
01Specs
| Publisher | poolside · poolside |
|---|---|
| Family / series | poolside Laguna · Laguna-XS.2 |
| Type | Text LLM (text-generation) |
| Parameters | 33B total · 3B active (MoE) Source: official model card“33B total with 3B activated per token” |
| Context length | 256K tokens (262,144)max_position_embeddings in config.json |
| Architecture | 40 layers (10 full-attention) · 30 sliding-window layers (512) · GQA · MoE |
| Released weights | BF16 · 66.9 GBdownload |
| Licence | Apache-2.0 · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed. |
| Released | 2026-04-23 (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) | 66.0 GB | 73.5 GB | 74.5 GB | 78.5 GB | RTX PRO 6000 96GB | Mac 128GB |
| FP8 / INT8 / Q8_0 | 35.1 GB | 39.5 GB | 40.5 GB | 44.5 GB | 2 × RTX 5090 (64GB) | Mac 64GB |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 20.0 GB | 22.9 GB | 23.9 GB | 27.9 GB | RTX 5090 32GB | Mac 64GB |
KV cache per request: 8K 0.4 GB · 32K 1.4 GB · 128K 5.4 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
- 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-xs.2
ollama run laguna-xs.204Official repositories & quantised builds
4 official repo(s) from poolside.
- Laguna-XS.2 main
- Laguna-XS.2-NVFP4 NVFP4
- Laguna-XS.2-FP8 FP8
- Laguna-XS.2-INT4 INT4
05Other poolside Laguna models
All 4 →| Model | Params | 4-bit total | Released |
|---|---|---|---|
| Laguna-S-2.1 | 118B / 8B | 79.7 GB | 2026-07-13 |
| Laguna-XS-2.1 | 33B / 3B | 22.9 GB | 2026-06-20 |
| Laguna-M.1 | 225B / 23B | 153 GB | 2026-06-15 |
06Sources
Data checked 2026-09-22. Parameters, context and licence are copied from the official repository; memory figures are our estimates.