Luis, Sara and Peter are the chatbot layer. Pick one from the sidebar and every reply takes on that voice: concise, friendly or technical. The model and the pipeline underneath stay the same.
A chatbot is the third of three layers. The model you pick does the thinking, the pipeline shapes how the answer comes out, and the chatbot sets the tone. It is one short instruction on top of everything else, so switching chatbots never changes what the model knows or how hard it works. It changes how the answer reads.
Which one you like is a matter of taste. Some people want the answer and nothing else. Some want to feel they are talking to someone. Some want the full technical picture every time. Pick the one you like talking to, and pick a different one for your next conversation whenever you feel like it.
The answer, and nothing else.
Luis gets to the point. Short replies, no warm-up, no recap at the end. When a yes or a number will do, that is what you get. Pick Luis when you know what you are asking and want to move on.
Warm, patient, on your side.
Sara talks to you like a person. She acknowledges what you are dealing with, explains without talking down, and checks that the answer actually helps. Pick Sara when the question is personal, when you are learning something new, or when a plain answer would land cold.
The full picture, with the reasons.
Peter gives you the mechanism, not just the result. Definitions are exact, edge cases get named, and trade-offs come with the why. Pick Peter when you want to understand how something works, not only what to do.
Each chatbot adds one short directive to the instructions the model receives: be concise, be friendly, be technical. It applies to every reply in the conversation, whichever model and pipeline you have picked.
The chatbot does not change which model thinks or how hard it works, and it does not change the pipeline. Pick a model for depth, a pipeline for the kind of answer, and a chatbot for the voice.
A conversation keeps the chatbot that first answered it, so a thread reads in one voice from start to finish. Pick a different chatbot and it applies from your next new conversation. Sign in and your pick is remembered.
Under all three, the same house rules keep every answer direct: answer first, no filler, prose first. Luis is not blunter about the facts and Sara does not soften them. The voice changes; the standard does not.
Luis, Sara and Peter started in October 2022 as Deep Spanish, our first AI experiment: three Spanish practice chatbots built on an early language model. Luis was the sarcastic one, Sara the super-friendly one, and Peter the grammar nerd.
Today they run on current open-weights reasoning models, answer in any language and on any topic, and keep a trace of their old personalities in the tone each one sets.
No. The model you pick does the thinking on every turn, and the pipeline shapes the answer. The chatbot adds one short directive on top: be concise, be friendly or be technical. Change the chatbot and the voice changes; the model and the pipeline stay exactly as you set them.
No. A conversation keeps the chatbot that first answered it, so it reads in one voice from start to finish. Start a new conversation to talk to a different chatbot. If you are signed in, your pick is remembered for new conversations.
Luis if you want the answer and nothing else. Sara if you want to feel you are talking to someone, or you are learning something new. Peter if you want the full technical picture with the reasons. There is no wrong choice, and you can pick a different one for your next conversation.
Not any more. They started in 2022 as Spanish practice chatbots, but today they answer in any language and on any topic. Write in Spanish and they answer in Spanish; write in English and they answer in English.
No. All three are Berges AI, running on the same open-weights reasoning models under the same house rules. The name and the tone are the only difference.