# AI Playground URL: https://discuvr.org/ai-playground/ Last updated: 2026-11-07 Send one prompt to up to 24 models from OpenAI, Anthropic, Google, Groq, Mistral, DeepSeek, xAI and more, side by side. Bring your own API keys. AI Playground sends one prompt to several models at once and streams every answer in its own column, with the time taken and the tokens used. Benchmarks rank models on someone else's questions; this shows how each model handles yours. ## What it does - 12 provider options: OpenAI, Anthropic, Google Gemini, Sarvam, Groq, Mistral, DeepSeek, OpenRouter, Together AI, Fireworks, xAI and any OpenAI-compatible endpoint. - Up to 24 models in one run, streamed in parallel. - One shared system message, plus temperature, max tokens and thinking level per model. - Attach a PDF, DOCX or TXT file (up to 5 MB and 50,000 characters of text) and every model gets the same text. - Follow-up prompts keep each model's own conversation. - Signed-in users can save a session and reopen it later. ## How it works 1. Sign in, open Keys and paste an API key for each provider you want to use. 2. Add models, write the system message once and type your prompt. 3. Run, then compare the answers, timings and token counts side by side. ## Where your keys and prompts go Keys are kept in your browser's local storage and sent with each run; the server uses them to call the provider and does not save them. Clear all keys on the Keys page removes them from the browser. Prompts and answers are not saved either, unless you click Save, which stores the system message, the model list and the conversations in your account until you delete the session. The usage record holds which providers ran and how many models, never the prompt or a key. ## Web search and the date With web search on, a prompt that looks time-sensitive (words such as latest, today or price, or a year such as 2026) triggers a DuckDuckGo search, and the top five results are added to every model's system message, so all models see the same sources. The date switch adds the current UTC date and time in the same way. ## What a fair comparison needs Keep the system message and the prompt identical across models and change one thing at a time. Run a prompt more than once before drawing a conclusion: a single sample from a sampled model is an anecdote. Set the same temperature where a model accepts one; some reasoning models do not. ## Limits Prompt and system message up to 20,000 characters each, 24 models per run, up to 40 earlier messages per model, and 30 runs per 10 minutes per IP address. A custom endpoint must be a public address: private, loopback and other internal ranges are refused. ## What an attached file gives the models The models get the file's text, not the file itself. A PDF gives only its text layer, so a scanned page comes through empty: there is no OCR. A DOCX gives its body paragraphs, without tables, headers, footers or footnotes. A TXT file is read as UTF-8, or as Latin-1 if that fails. The text is added to every model's system message and stays attached for follow-up prompts until you remove it, so its tokens are billed on every run. Text over 50,000 characters is refused when you run. ## Questions **Do you store my API keys?** Not on the server. They sit in your browser's local storage, so they are still there on your next visit in the same browser, and travel with each run to call the provider. Clear all keys on the Keys page removes them. **Who pays for the tokens?** You do, directly to each provider, at their rates. The playground adds no cost of its own. **Which providers are supported?** OpenAI, Anthropic, Google Gemini, Sarvam, Groq, Mistral, DeepSeek, OpenRouter, Together AI, Fireworks, xAI, and any endpoint that speaks the OpenAI chat completions format. **Do I need an account?** Yes. Running models needs a free sign-in, which keeps the endpoint from being abused. Saving sessions uses the same account. **Why do the models not see the table in my Word file?** Only the body paragraphs of a DOCX are extracted; tables, headers, footers and footnotes are left out. Paste the table into the prompt as text instead. The character count next to the file name shows how much text was extracted. **Can I connect a local model such as Ollama or LM Studio?** Only if it is reachable from the internet. Choose Custom (OpenAI-compatible) and enter the full chat completions URL, for example https://your-host/v1/chat/completions. The request is sent from the Discuvr server, not from your browser, and any address that resolves to localhost, a private network or another reserved range is refused, so http://localhost:11434 (Ollama) or http://localhost:1234 (LM Studio) will not work. Put the server behind a public HTTPS address you control, and save a key for Custom on the Keys page even if your server ignores it, because a run without a key stops with an error. --- Part of Discuvr (https://discuvr.org/): Discuvr is a collection of free, focused tools for day-to-day professional work - office and data utilities, legal and compliance helpers, privacy and security checks, and developer formatters. Every tool has its own page, and most can also be called from ChatGPT or Claude over MCP.