Arbiter House

Find the right model for your prompt. Arbiter House helps you explore which language models may suit your task. Write or paste a prompt, compare matches across a bundled catalog of 458 models, and explore ranked results on iPhone and iPad. COMPARE TASK FIT Each model receives an estimated match score based on your prompt and its published description. Explore exact model IDs, names, logos, and descriptions, then copy a model ID to use elsewhere. CHECK A MESSAGE Prompt Check reviews text against 30 transparent signals, including sensitive information, instruction conflicts, model-training requests, permissions, and content that may need closer attention. It highlights signals to review; it is not a security audit, compliance decision, or guarantee about how any model will behave. ROOM FOR YOUR IDEAS Use a multiline editor, resize it, or expand it to full screen. Long prompts can be shortened for model scoring while the original stays in your editor. Follow progress, cancel unfinished work, and retry remaining models. REVISIT YOUR RESULTS Save your latest 100 scored comparisons on this device, including clearly marked partial results. Open saved history offline, reuse a prompt, delete individual runs, or clear all local data in Settings. Prompt Check reports stay in the current app session and are not added to History. UNDERSTAND WHAT YOU SHARE Before remote model scoring or a Prompt Check, choose whether to allow it. If you allow scoring, your prompt excerpt and catalog model information are sent over HTTPS to the independent typesafe.pro gateway and TypeSafe AI. A Prompt Check sends the text you entered and its fixed evaluation criteria. Retries may send submitted text again. Shortening for model scoring does not remove personal information. The providers may retain submitted content and network or diagnostic records under their policies. Remote checks and scoring are not zero-retention services. Please avoid entering passwords, API keys, confidential material, or other sensitive information. Fresh checks require an internet connection and depend on service availability. No account, subscription, purchase, or user API key is required. Your draft and saved comparison history are stored on this device. Clearing local data removes those local records and cancels unfinished scoring; it cannot retract requests already sent or delete the providers' records. The Privacy Policy explains processing, retention, and how to make a privacy request. Privacy and support links are available in Settings. Match scores are estimates from model descriptions, not benchmarks or guarantees of quality, cost, speed, or current availability. The bundled catalog is a dated snapshot. Arbiter House recommends models; it does not run your prompt on the recommended models.