olace

[ the app ]

Models

Olace treats every model you can reach as one library: models on this device, models on your paired devices, Olace Cloud models, and models behind your own API keys. Pick any of them from the model selector in a chat.

Local models

Local models run through Ollama, LM Studio, or llama.cpp on a computer. Install and manage them in the Model hub (Settings › Local AI › Model hub):

  • Browse a curated catalog with search and capability filters (Reasoning, Vision).
  • Models that need more GPU memory than you have are labeled before you download.
  • Delete a model any time; you can download it again later.
  • Unload models frees GPU memory immediately; models reload on next use.

The same works from the terminal:

olace list
olace pull qwen3.5:4b
olace rm qwen3.5:4b
olace unload

Paired models

Models on your other devices appear in the same selector, marked with the device they run on. Chats with them stream over an end-to-end encrypted connection. You can also install models on a paired computer remotely, from the Model hub or with olace pull <tag> --device <name>. See Remote Access.

Olace Cloud

Olace Cloud offers top open models with zero data retention, for when your host machine is unavailable. Cloud turns are metered in credits. Providers are required to hold nothing and train on nothing; if a model is not available under zero data retention, Olace tells you instead of quietly relaxing the rule.

Bring your own key

On Plus and up, connect your own provider keys under Settings › Advanced › Keys & routing. Supported providers: OpenRouter, Ollama (ollama.com), Together AI, Moonshot Kimi, and Z.AI GLM.

  • Keys stay on your device and are used directly from it. With sync on, keys are encrypted before syncing to your other devices, unreadable to Olace.
  • Cloud turns on your key never spend your credits.
  • On OpenRouter, Olace requests zero data retention and blocks data collection where the provider supports it. For other providers, retention follows your account settings with them.

Model names in the CLI and Bridge

Wherever you type a model id, the same forms work:

ollama/qwen3.5:4b               local Ollama
lmstudio/google/gemma-4-e4b     local LM Studio
llamacpp/qwen3.5:4b             local llama.cpp
home-pc/ollama/qwen3.5:14b      a paired host named "Home PC"
moonshotai/kimi-k2.6            Olace Cloud

Unambiguous shorthand is accepted, so qwen3.5:4b works when only one provider has it.