Documentation
Everything you need to get started with ThinkHere.
Getting Started
ThinkHere runs entirely in your browser. There is nothing to install, no API keys to configure, and no account required to start chatting.
The first load downloads model weights (1–3 GB depending on the model). After that, the model loads from your browser's cache in seconds.
Feature Tiers
ThinkHere offers three tiers. All tiers run the model locally in your browser — we never see your conversations.
| Feature | Free | Logged In (Free) | Paid (Monthly) |
|---|---|---|---|
| Gemma 3n E2B | ✓ | ✓ | ✓ |
| All models (9+) | — | ✓ | ✓ |
| Multimodal (images) | ✓ | ✓ | ✓ |
| System prompts | — | ✓ | ✓ |
| Temperature / generation controls | — | ✓ | ✓ |
| Conversation history | — | ✓ | ✓ |
| Export conversations | — | ✓ | ✓ |
| Knowledge base / RAG | — | ✓ | ✓ |
| File upload as context | — | ✓ | ✓ |
| Priority support | — | — | ✓ |
| Premium features (coming soon) | — | — | ✓ |
Device Requirements
ThinkHere runs AI inference on your local hardware, so device capability matters.
Available Models
The following models are available in the logged-in tier. The free tier includes Gemma 3n E2B only.
| Model | Family | Size | Tier |
|---|---|---|---|
| Gemma 3n E2B | Gemma | ~3 GB | Free |
| Gemma 3n E4B | Gemma | ~4 GB | Logged In |
| Gemma 3 1B | Gemma | ~1.3 GB | Logged In |
| Gemma 2 2B | Gemma | ~1.6 GB | Logged In |
| Phi-4 Mini | Phi | ~2.3 GB | Logged In |
| Llama 3.2 1B | Llama | ~1.3 GB | Logged In |
| Llama 3.2 3B | Llama | ~2.1 GB | Logged In |
| Qwen3.5 0.6B | Qwen | ~0.8 GB | Logged In |
| Qwen3.5 1.7B | Qwen | ~1.4 GB | Logged In |
Model sizes are approximate and represent the download size. Larger models generally produce higher quality responses but require more RAM and may generate tokens more slowly.