Poolside

Laguna S 2.1

Name: Laguna S 2.1
Rating: 88 (29 reviews)
Author: Poolside

Frontier

HuggingFace

45.3KDownloads627CurtidasJul 2026Publicado1.0M tokensContextoOpenMDW 1.1Licença83 ForteQualidade

Laguna S 2.1 (117.5999984741211B parameters) requires approximately 74.0 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 8B active parameters, it uses less memory than its total parameter count suggests. For the best balance of quality and speed, we recommend hardware with at least 86 GB of VRAM.

Comece agora

— copie e cole para rodar localmente

Copy-paste commands to run Laguna S 2.1 on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
  --hf-repo "poolside/Laguna-S-2.1" \
  --hf-file "Laguna-S-2.1-Q4_K_M.gguf" \
  -c 4096 -ngl 99

Quick specs

Parameters117.6B (8B active)

Architecturemoe (MoE)

Context1.0M tokens

Modalitytext

Min RAM45.9 GB

Rec. RAM71.7 GB (Q4_K_M)

LicenseOpenMDW 1.1

FamilyLaguna

✓ Code✓ Reasoning

About this model

Laguna S 2.1 is Poolside's 118B-total Mixture-of-Experts model with ~8B parameters activated per token, built for agentic coding and long-horizon work. It uses 256 routed experts plus one shared expert with a softplus token-choice router, grouped-query attention, and a 1:3 global-to-sliding-window attention layout across 48 layers, with a 1M-token context.

•118B total / ~8B active per token — 256 routed experts plus 1 shared expert.
•1,048,576-token context window for long-horizon agentic coding.
•Hybrid attention: 12 global layers + 36 sliding-window layers (window 512), cutting KV-cache cost.
•Native interleaved reasoning between tool calls, plus a DFlash draft model for speculative decoding.
•Official FP8, NVFP4, INT4 and GGUF variants published by Poolside.

Modelos relacionados

Escolhas rápidas

Melhor custo-benefícioA

MacBook Pro M3 Max 128GB~$2,499 — 16 tok/s

Melhor no geralS

AMD Instinct MI250X 128GB~$15,000 — 106 tok/s

Melhor hardware

Melhores opções para Laguna S 2.1

AMD Instinct MI250X 128GBS

128 GB

AMD Instinct MI300A 128GBS

128 GB

AMD Instinct MI250 128GBS

128 GB

Gaudi 3 128GBS

128 GB

Intel Data Center GPU Max 1550 128GBS

128 GB

Rodar este modelo

Laguna S 2.1 on AMD Instinct MI250X 128GB Laguna S 2.1 on AMD Instinct MI300A 128GB Laguna S 2.1 on AMD Instinct MI250 128GB

Opções de quantização

Estimativas de VRAM por nível de quantização

No hardware detected — fit column shows raw VRAM estimates

Quant	Bits	VRAM	Quality	Fit
Q1_0_G128	1.125	16.9 GB	Very Low	—
Q2_0_G128	1.71	31.4 GB	Low	—
Q2_K	2	45.9 GB	Low	—
Q3_K_S	3	57.6 GB	Low	—
NVFP4	4	65.9 GB	Medium	—
Q4_K_M	4	71.7 GB	Medium	—
Q5_K_M	5	84.7 GB	High	—
Q6_K	6	96.4 GB	High	—
Q8_0	8	125.8 GB	Very High	—
F16	16	241.1 GB	Maximum	—

Compatibilidade de hardware

Estimativas de compatibilidade para todo o hardware

Abrir calculadora

Computing compatibility...

Detalhamento de memória

Reference: RTX 2060 6GB

Weights71.7 GB

KV Cache0.7 GB

Runtime0.9 GB

Headroom0.6 GB

Perguntas frequentes

FAQ — Laguna S 2.1

Veja também

Guia de Quantização Metodologia de Pontuação Abrir calculadora