In a nutshell: Tresor offers several AI models. Use Instant mode for fast, direct answers, or Thinking mode for problems that require deep reasoning. All models run inside secure enclaves.
Tresor currently offers these AI models, all running inside confidential computing environments:
Model | Strengths |
|---|---|
Kimi K2.5 (Recommended) | Versatile, strong reasoning, vision capable |
GPT-OSS 120B | Large, powerful general-purpose model |
DeepSeek v3.2 | Fast, efficient, great for drafting |
Llama 3.3 70B | Open-source, strong at following instructions |
Mistral 24B | Fast and uncensored |
GLM 5 | Large context window, multilingual |
All models support Thinking mode and image understanding. Tresor handles these capabilities at the platform level, so you don't need to worry about which model supports what — just pick the one that fits your task.
💡 Tip: If you're not sure which model to use, start with Kimi K2.5 — it's Tresor's recommended model and handles most tasks well.
Models that support reasoning have two modes:
Best for: Quick questions, drafting, summarizing, translating, everyday tasks.
How it works: The model responds directly and quickly.
Example: "Translate this email to French" or "Summarize this paragraph."
Best for: Complex analysis, multi-step problems, strategic planning, legal reasoning, math.
How it works: The model takes extra time to reason through the problem step by step before answering. You'll see a "Thinking…" indicator while it works.
Example: "Analyze the tax implications of restructuring our subsidiary" or "Find the logical flaw in this argument."
You can use Thinking mode with any model.
Look at the top navigation bar of your chat.
Click the model selector (it shows the name of the currently active model).
Choose your preferred model from the dropdown.
If the model supports it, toggle between Instant and Thinking mode.
[Screenshot: The model selector dropdown in the top navigation showing available models and the Instant/Thinking mode toggle]
You can change your model at any time, even in the middle of a conversation. The new model applies to your next message.
Here are some rules of thumb:
| Task | Recommended model | Mode | | ------------------------------------- | -------------------------- | -------- | | Quick Q&A, drafting, emails | DeepSeek v3.2 or Kimi K2.5 | Instant | | Complex reasoning, analysis | GPT-OSS 120B or Kimi K2.5 | Thinking | | Analyzing images or PDFs with visuals | Kimi K2.5 | Either | | Following detailed instructions | Llama 3.3 70B | Instant | | Unfiltered / uncensored responses | Mistral 24B | Instant | | Multilingual tasks | GLM 5 or Kimi K2.5 | Either |
All models run inside secure enclaves. Your privacy is the same regardless of which model you choose.
Model availability can occasionally change. If a model is temporarily unavailable, Tresor will let you know — simply switch to another.
You don't need to pick the "right" model to get started. Try a few and see which one suits your work best. You can always switch.
Thinking mode takes longer but produces more thorough answers. If you need speed, use Instant mode.
🔒 Privacy: Every model listed in Tresor runs inside a verified confidential computing environment. Your messages never leave the sealed enclave, regardless of which model you choose. The providers themselves (OpenAI, Meta, etc.) do not receive your messages directly — Tresor's enclave infrastructure processes them.
Your first conversation — Send your first private message.
AI models and Thinking mode — Change models during a conversation.
Verifying your conversation's privacy — See the proof for yourself.
