TuttiFruttiThis is a translation for your convenience. The Spanish version is the binding one and prevails in case of any difference.
Transparency about artificial intelligence
1. You're talking to an artificial intelligence
At TuttiFrutti every reply in the chat is written by an artificial intelligence system, not a person. Nobody on our team reads or reviews your conversations before you get the reply. We tell you this in accordance with art. 50 of the European AI Regulation.
2. Which models answer and from which provider
TuttiFrutti has no models of its own: it uses other companies' models, through their providers. Your message and that chat's earlier conversation are sent to the provider that answers.
| In the app | Model | Who runs it | Where |
|---|---|---|---|
| TuttiFrutti (free) Tried in this order while there's free quota left for the day |
1. GPT-OSS 120B (OpenAI's open model) | Groq | USA |
| 2. Qwen 3.8 27B (Alibaba's open model) | Groq | USA | |
| 3. Gemma 4 26B (Google's open model) | Cloudflare Workers AI | USA | |
| 4. DeepSeek Flash, when the previous ones run out | DeepSeek | China | |
| TuttiFrutti: backup Only if DeepSeek can't answer |
5. DeepSeek V4 Flash | OpenRouter, which sends it to a company that runs that model and doesn't keep the messages | USA and that company's country |
| 6. GPT-5 nano | OpenAI | USA | |
| TuttiFrutti Pro | DeepSeek V4 Pro | DeepSeek | China |
| TuttiFrutti Pro: backup Only if DeepSeek can't answer, in this order |
1. DeepSeek V4 Pro | OpenRouter, which sends it to a company that runs that model and doesn't keep the messages | USA and that company's country |
| 2. GPT-5 mini | OpenAI | USA | |
| 3. DeepSeek V4 Flash | OpenRouter | USA and that company's country | |
| Models in the developer API | DeepSeek Flash and DeepSeek V4 Pro (no backup) | DeepSeek | China |
| Images and reading aloud | FLUX.1 schnell (images) and MeloTTS (voice) | Cloudflare Workers AI | USA |
| OpenAI's image and voice models, if Cloudflare can't handle it | OpenAI | USA |
Under each reply in the chat we state that it was generated by an AI and which specific model from this list wrote it, with its provider. We may change the models and providers; when we do, we'll update this page.
3. What you should know about their replies
- They can be wrong. Language models generate likely text, not checked text. They can make up facts, quotes or sources, and get calculations or code wrong.
- They can be out of date. Each model knows the world up to its training date and doesn't search the internet.
- They can have biases from the data they were trained on.
- They aren't professional advice. They don't replace a doctor, lawyer, financial adviser or any other professional.
- Each model has its provider's safety filters, which may make it refuse some requests.
4. Your data and the models
We don't train models or keep your conversations on the server. What each provider does with the messages it receives is set by its terms; we explain it in the privacy policy. Don't write sensitive personal data or other people's data in the chat.
5. AI-generated content
If you publish or share texts generated with TuttiFrutti, bear in mind that in some cases the law requires stating that they were generated by AI (for example, texts published to inform the public on matters of public interest, if they haven't had human review or editorial responsibility; art. 50.4 of the AI Regulation).
The images and audio TuttiFrutti creates are generated by an AI. Today they don't yet carry a mark that allows this to be detected automatically; we'll add it. In the meantime, if you publish them, say they were made with AI.
Whether TuttiFrutti, by offering the chat under its own brand, counts as a “provider” of an AI system (as well as using third-party models). If so, the machine-readable marking of art. 50.2 (synthetic images, audio, video and text) is TuttiFrutti's responsibility, and the image and voice features are already live without that marking: it's urgent (for example, C2PA or IPTC metadata in the images). Also review the AI literacy obligation of art. 4 for anyone working on the service.
Using the chat doesn't seem to be high-risk (annex III) or a prohibited practice (art. 5), as long as it isn't offered for purposes such as recruitment, creditworthiness assessment or education with grading. If tools for those purposes are launched, review again.
6. Contact
If a reply seems harmful, illegal or misleading to you, or you have questions about how we use AI, write to us at mindrapeofficial@gmail.com.