← ReverseML Mini

Expert Modes

Modes of one Mini Core — not separate products. You pick a mode in the app (⚙️); the base model is shared, only the specialization changes.

Honesty note: trained = a mode with its own fine-tuned weights (downloadable GGUF). profile = the same base + domain system-prompt, tools, references and its own memory (no separate weights).

Two kinds of expertise

ReverseML Agent (online) — deep technical research, documents, code and calculations in the cloud Workspace. Open Workspace →

ReverseML Mini (on-device) — the local modes below, running privately on your own device.

🛠️ Engineer · specialized trained model
⬇️ Download weights (Ollama)
Manuals, sensor logs, calculations (Kalman/ODE) — answers with cited sources. Has its own LoRA-trained weights over the base.
💻 Developer / Coder · domain-adapted profile
A private local coding helper — minimal, correct code, offline.
⚗️ Chemist · domain-adapted profile
Exact equation balancing (deterministic) + chemistry Q&A.
📚 Teacher · domain-adapted profile
A patient step-by-step tutor, offline.
⚖️ Legal Research Mode · domain-adapted profile
General legal information and research assistance.
For informational and research assistance. Not a substitute for qualified professional advice.
🩺 Medical Information Mode · domain-adapted profile
General health information and research assistance.
For informational and research assistance. Not a substitute for qualified professional advice.
🦅 General · base
The shared base — everyday assistant for phones, laptops and tablets, fully offline.
Open the app to pick a mode →

On phones/tablets a mode = the shared base + its focus and memory. On desktop, some modes (e.g. Engineer) also ship their own trained weights; more roll out as nightly training completes.