Query 50+ AI models in parallel with one prompt and see which one scores highest automatically.
Test how AI models respond to text encoded in leetspeak, Morse code, braille, or Unicode substitutions.
Run five model-prompt combinations in parallel and get the best response for red-teaming research.
Automatically tune AI generation parameters using a thumbs-up/thumbs-down feedback loop over time.
| elder-plinius/g0dm0d3 | a16z-infra/companion-app | craft-ai-agents/craft-agents-oss | |
|---|---|---|---|
| Stars | 5,956 | 5,954 | 5,960 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | easy | moderate | easy |
| Complexity | 1/5 | 3/5 | 3/5 |
| Audience | researcher | developer | vibe coder |
Figures from each repo's GitHub metadata at analysis time.
Requires a paid OpenRouter API key to connect to AI models, no install or build step needed, just open index.html in a browser.
G0DM0D3 is an open-source browser-based chat interface for interacting with over 55 AI language models through a single web page. It connects to models including Claude, GPT, Gemini, Grok, Mistral, LLaMA, and DeepSeek via the OpenRouter API service, which routes requests to multiple AI providers. Users supply their own OpenRouter API key. The application centers on several distinct modes. GODMODE CLASSIC runs five model-and-prompt combinations in parallel and returns the best response. Each combination pairs a specific model with a pre-built prompt designed to push past content restrictions, which the project describes as jailbreaking for red teaming research. ULTRAPLINIAN is a multi-model evaluation engine that queries up to 51 models simultaneously, scores their responses on a composite metric, and surfaces the highest-scoring result. A feature called Parseltongue applies text transformations to prompts before sending them, such as converting text to leetspeak, Morse code, braille encoding, or Unicode character substitution. The project frames this as a tool for studying how AI models respond to differently encoded inputs. AutoTune automatically selects generation parameters such as temperature based on the query type, with a feedback loop that refines its choices over time based on thumbs-up and thumbs-down ratings. The application is a single index.html file with no build process or framework dependencies. It runs entirely in the browser. Chat history and the API key are stored in browser localStorage only, never transmitted to the application's servers. Telemetry is anonymous and can be turned off in settings. The project is licensed under AGPL-3.0. An optional self-hosted API server adds an opt-in mode to contribute anonymized chat data to a public research dataset on HuggingFace.
G0DM0D3 is a browser-based chat interface that lets you query 55+ AI models through a single webpage using your OpenRouter API key. It includes modes that run multiple models in parallel, score responses, and apply text encodings to test how different AI models behave.
Mainly TypeScript. The stack also includes TypeScript, HTML, JavaScript.
You can use, study, and share this freely, but if you distribute any modified version, including running it as a hosted service, you must release your changes under the same AGPL-3.0 license.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
Verify against the repo before relying on details.