Laguna S 2.1
Neural Network
Laguna S 2.1 is a coding agent by Poolside: a 118B parameter model with 8B active for agentic development tasks.
Max answer length
(in tokens)
Context size
(in tokens)
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Laguna S 2.1?
Laguna S 2.1 is a Poolside model focused on coding in agentic mode: it is designed for long chains of steps where you need to read a project, make edits, and verify results. Architecturally, it has 118 billion parameters, with 8 billion active per request, and scores 70.2% on Terminal-Bench 2.1, which tests the ability to complete tasks in a terminal. In BotHub, the model is accessible from Russia without a VPN or foreign card: pay with a Russian card in rubles only for tokens actually consumed, and unused ones do not expire. With over 250 models in one window, it's easy to compare results. Requests are encrypted via AES-GCM and not saved, and a unified OpenAI-compatible API allows you to switch models without rewriting integrations. For teams, we offer contracts, invoices, and an admin panel with limits. It is useful for writing features, cleaning up legacy code, debugging failing tests, automating CI/terminal routines, analyzing repositories before reviews, or building your own developer agent via API.Frequently asked questions about Laguna S 2.1
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
laguna-s-2.1 by poolside works with text, documents, and images: it analyzes large codebases, helps with refactoring and debugging, solves reasoning tasks, and calls external functions. A context window of up to 1,048,576 tokens allows you to keep an entire project in one dialogue.
Up to 131,072 tokens per response. This is enough to get an entire code module, a detailed document analysis, or a long technical text without splitting it into parts. If the response cuts off, just ask it to continue from where it stopped.
Yes, the reasoning mode is confirmed: the model first outlines the solution path and only then provides an answer. This is especially noticeable in multi-step tasks like debugging, math, and analyzing long documents. For simple queries, the difference is smaller, but the answers remain accurate.
Yes. The model supports function calling and structured JSON output, making it easy to integrate into agents and pipelines: it selects the tool itself and returns data in the specified schema. In BotHub, this works via a unified OpenAI-compatible API.
You can input text, images, and documents—for example, an interface screenshot, an architecture diagram, or a PDF with specifications. Audio and video are not listed among the supported input formats in our data. The model returns results as text.
There is no specific note about Russian language support in our data, so we cannot guarantee accuracy—please test it with your tasks. If the result is unsatisfactory, you can switch to another model in BotHub with one click without rewriting your integration.