Polyglot Studio: A Local-First LLM Client Supporting 360+ Models with Multi-Model Parallel Comparison

Polyglot Studio is an open-source client application for LLM enthusiasts, developed and maintained by devrathc. The source code is hosted on GitHub (link), and it was released on May 23, 2026. Core Features:

• Built with Next.js, using local-first + BYOK (Bring Your Own Key) model to ensure privacy
• Supports over 360 models via OpenRouter API (including GPT-4, Claude, Gemini, Llama, etc.)
• Allows parallel comparison of up to 5 models, enabling intuitive comparison of answer quality and style
• Provides features like free routing, multimodal dialogue, and conversation history management

Original Author and Source

• Original Author/Maintainer: devrathc
• Source Platform: GitHub
• Original Link: https://github.com/devrathc/polyglot-studio
• Release Date: 2026-05-23

Project Overview

Polyglot Studio is an open-source LLM client designed to break the limitations of single models, allowing users to access over 360 models while maintaining control over their data. Its core value lies in the multi-model parallel comparison feature.

Local-first

Data is stored in the local browser or self-hosted environment, ensuring the privacy of conversation history. It is suitable for privacy-sensitive users and enterprises.

BYOK (Bring Your Own Key) Mode

Users need to configure their own OpenRouter API key:

• Transparent Costs: Pay directly to OpenRouter
• Model Freedom: Access over 360 mainstream models
• Privacy Protection: Keys and request content are only transmitted during communication; the application does not store or intercept them.

Multi-Model Parallel Comparison

Open up to 5 conversation windows simultaneously, suitable for model selection, quality evaluation, cost optimization, and multi-perspective analysis.

Support for 360+ Models

Covers OpenAI (GPT-4, etc.), Anthropic (Claude series), Google (Gemini series), open-source models (Llama3, etc.), and vertical domain professional models.

Other Features

• Free Routing: OpenRouter automatically selects the optimal model
• Multimodal Support: Upload images/documents to process visual content
• Conversation Management: Create/archive/search/export conversation records.

Frontend Framework

Built with Next.js App Router, using a hybrid architecture of React Server/Client Components:

• Excellent development experience (TypeScript, hot reloading)
• Performance optimization (reduces client-side JS size)
• Flexible deployment (Vercel, self-hosted, or static export)

Local Storage

Uses IndexedDB/localStorage for data storage; self-hosted versions can configure server-side persistence.

API Integration

Communicates with OpenRouter via REST API, supports streaming responses, and handles error retries, rate limits, and other logic.

• AI Researchers/Developers: Efficiently test the performance of different models
• Content Creators: Use multi-model comparison to refine manuscripts
• Enterprise Model Selection: Conduct POC tests to evaluate model performance and costs
• Privacy-Sensitive Users: Local-first approach ensures data security.

Contribution and Learning

Provides CONTRIBUTING.md (contribution guide) and LEARNING.md (learning resources). The code is open-source under the MIT license.

Summary

Polyglot Studio focuses on open API frontend, local-first privacy, and multi-model support, providing users with flexibility, control, and privacy protection. It is an innovative choice for LLM clients.