The models
Core 3.1, Core 4, Logos 2 and Logos 3.5 side by side — what each is for and how to switch.
Socria ships four models behind one switcher, and they are not four sizes of the same thing — they are different amounts of machinery around the conversation. Core 3.1 is a voice with a memory and a running read of the thread. Core 4 is a voice that decides, each turn, which part of the work is yours. Logos 2 is an environment: the conversation beside a live map of your reasoning, with models the engine builds inside it. Logos 3 is that environment with a second seat.
Side by side#
| Core 3.1 | Core 4 | Logos 2 | Logos 3.5 | |
|---|---|---|---|---|
| In a sentence | Assertive pattern-naming with thread memory and adjustable depth | Contributes by default; holds work back only on a reason you gave | The conversation plus a live map of your reasoning, and models the engine builds, which you move, read, edit and undo | Logos 2 for two people at once, with Socria as the layer between them |
| Account | Not for a first session, kept in your browser; then sign-in. Free: 4 new chats and 20 messages a day, shared with Core 4 | Sign-in (or an access key). Free: 4 new chats and 20 messages a day, shared with Core 3.1 | Sign-in (or an access key). Free: 2 lines of thinking a month and 10 messages a day, shared with Logos 3, then Socria One | Sign-in, for both people. Free: counted with Logos 2 |
| Thinking depth | All four modes | None, deliberately — it judges depth itself, and gives you readability and length instead | None — no depth modes. It answers at one depth and puts the effort into the map and the model | All four modes, in the model picker |
| Memory | Thread memory, syntheses, insights, a cross-conversation journey | A record of the reasoning itself — every claim, objection and decision with whose it is, enforced in code | The map, the models and their revisions — a model keeps its history, so you can undo an edit rather than regenerate a picture | The room keeps its own record, each line with who wrote it; a shared room is saved to neither account and never feeds memory |
| Extra surfaces | Synthesis & insight cards, choice chips | Projects, source cards, exact arithmetic, attachments | Thinking Map, Board, plots, Draft Space, Research, and the model workspace: surfaces, mechanisms, systems of equations, fitted specifications and three simulations — a black hole, the expanding universe, an orbit | All of Logos 2, plus the room: an invite, two seats, every line and every idea marked with whose it is |
| Writes prose for you | Only refining material you brought | Yes, unless the authorship is the point — and it says so when it stops | Never. It may propose a model, and the engine decides whether it computes — a proposal is never presented as a result | Never — and between two people it never takes a side either |
Switching between them#
The model picker sits bottom-right, beside the chat box. Picking Logos 2 does not navigate anywhere: the whole surface swaps in place inside /chat, because Logos 2 is a model, not a destination — and leaving it returns you to whichever Core model you were on before, not to a default.
Your choice is remembered per browser. In the sidebar, chats and lines of thinking sit in one list ordered by when you last touched each; lines of thinking carry a small map mark, and opening one switches you into Logos 2. The way back is the Socria chat button in the Logos 2 header — Chat on a narrower screen — which returns you to whichever Core model you were on.
Which one, when#
- Core 3.1 — a thread you will return to, or no account at all. It notices your language, names patterns without hedging, asks at most one question per turn, and periodically hands you a structured synthesis of what you have actually worked out.
- Core 4 — long, consequential work you will come back to: a decision that lives over weeks, a piece of research, a plan whose load-bearing assumption is worth finding. It contributes by default, almost never asks a question you do not need, and does not re-raise what you have already settled. It is the wrong tool for a quick answer, and it starts flat: on a first message there is nothing yet to be continuous with.
- Logos 2 — thinking with structure, or thinking about a system: decisions, learning, math, and anything you would rather move than read about — a surface, a mechanism, a set of differential equations, a model fitted to data, a black hole.
- Logos 3.5 — the same, with the everyday displays (charts, plans, comparisons, decisions — what is new) and with somebody else: a decision two people share, a disagreement worth making precise, a plan you are building together.