Laguna S 2.1

Laguna S 2.1 is Poolside's open-weight 118B-A8B Mixture-of-Experts model for agentic coding, long-horizon work, and million-token contexts.

Platform: Replicate

Language ModelOpen WeightsCoding AssistantAgent WorkflowsLong ContextMixture of ExpertsPoolside

Self-hosted BF16 or quantized weights / OpenRouter / Baseten / Vercel AI Gateway / Poolside chat

Open weights for self-hosting; Poolside reports an OpenRouter dedicated 1M-context endpoint at $0.10 input, $0.20 output, and $0.01 cache-read per million tokens

Commercial

🚀Function Overview

Laguna S 2.1 combines 118 billion total parameters, about 8 billion activated parameters per token, a 1,048,576-token context window, preserved reasoning between tool calls, and official full-precision and quantized weights.

Key Features

Official Hugging Face model ID poolside/Laguna-S-2.1
118B total parameters with about 8B activated per token
256 routed experts using top-10 routing plus one shared expert
1,048,576-token context window
Native interleaved reasoning controlled with enable_thinking
OpenMDW-1.1 license for commercial and non-commercial use subject to its terms
Official BF16, FP8, NVFP4, INT4, GGUF, MLX, and DFlash variants
Serving guidance for Transformers, vLLM, SGLang, TRT-LLM, and llama.cpp
Published limitations for tool-schema adherence, nested JSON tool calls, and overthinking

Use Cases

•Long-horizon repository work through a tool-using coding agent
•Self-hosted software-engineering and terminal-task evaluation
•Large-context codebase and document analysis
•Local or private inference with an appropriate quantized checkpoint
•Hosted model routing tests where provider context, price, and tool behavior are verified

⚙️Input Parameters

model

string

Use poolside/Laguna-S-2.1 for the official Hugging Face checkpoint or the exact provider model ID documented by a hosted surface.

messages

array

Chat messages for text, reasoning, coding, and tool-use workflows.

tools

array

Optional tool definitions; validate the selected runtime's Poolside tool parser and retry behavior.

enable_thinking

boolean

Chat-template control for reasoning; Poolside recommends preserving prior reasoning content in agent histories.

💡Usage Examples

Example 1

Input Parameters

{
  "model": "poolside/Laguna-S-2.1",
  "messages": [
    {
      "role": "user",
      "content": "Inspect this repository, reproduce the failing test, implement the smallest fix, and rerun the affected suite."
    }
  ],
  "enable_thinking": true
}

Output Results

Laguna S 2.1 returns text and optional tool calls through the selected serving runtime. Validate parser compatibility, preserved reasoning, test evidence, latency, and memory use before production adoption.

Quick Actions

Use NowView Documentation

Technical Specifications

Hardware Type: Self-hosted BF16 or quantized weights / OpenRouter / Baseten / Vercel AI Gateway / Poolside chat
Commercial Use: Supported
Pricing: Open weights for self-hosting; Poolside reports an OpenRouter dedicated 1M-context endpoint at $0.10 input, $0.20 output, and $0.01 cache-read per million tokens
Platform: Replicate

Related Keywords

Laguna S 2.1poolside Laguna S 2.1poolside/Laguna-S-2.1Laguna S 2.1 modelLaguna S 2.1 context windowLaguna S 2.1 licenseLaguna S 2.1 GGUFLaguna S 2.1 localLaguna S 2.1 OpenRouterLaguna S 2.1 coding model

Related Models

Bielik 1.5B v3 Instruct

Bielik-1.5B-v3-Instruct is a generative text model featuring 1.6 billion parameters. It is result of collaboration between the open-science/open-souce project SpeakLeash and the High Performance Computing (HPC)

Cordia-A6 Text Generation Model

A model for generating text sequences based on input prompts and adjustable parameters.

Claude Sonnet 4

Claude Sonnet 4 is a significant upgrade to 3.7, delivering superior coding and reasoning while responding more precisely to your instructions