Chutes

Chutes hosts leading open-source AI models with confidential compute (TEE) backing. Toby connects to Chutes via its OpenAI-compatible endpoint, so chat, streaming, tool calling, and structured outputs all work out of the box.

Personas that use the chutes provider send requests to https://llm.chutes.ai/v1 with slash-delimited model ids, for example deepseek-ai/DeepSeek-V3.2-TEE or Qwen/Qwen3-32B-TEE.

Get an API key

1. Sign in at chutes.ai.
2. Open Auth → Start to create an API key (starts with cpk_).
3. Copy the key.

Store the key in Toby, not in shell profiles or committed files.

Configure in Toby

Open Toby.app → Settings → AI → Chutes → API Key and paste your key.

Toby writes it to ~/.toby/credentials.json.

You can also set CHUTES_API_KEY in your environment instead of storing a key in credentials.

For a persona, set AI Provider to chutes, then pick a model from the list or enter a custom model id.

Model catalog

Chutes exposes a public model catalog at GET /v1/models. Toby fetches the live list automatically, even before you configure an API key, so you can browse available models in the persona editor without signing up first.

If the catalog cannot be reached, Toby falls back to a curated default list.

Recommended models

These models are available on Chutes with TEE (confidential compute) backing. All report confidential_compute=true.

Everyday chat

Model	Role
deepseek-ai/DeepSeek-V3.2-TEE	Strong all-rounder for chat, reasoning, and tool use. Default auxiliary model for Chutes.
Qwen/Qwen3-32B-TEE	Cost-aware model good for tool loops and everyday chat.
MiniMaxAI/MiniMax-M2.5-TEE	Balanced cost and quality for agentic workflows.

Long context

Model	Role
zai-org/GLM-5.2-TEE	1M token context window for long-context work.
zai-org/GLM-5.1-TEE	198K context for long documents.
moonshotai/Kimi-K2.6-TEE	256K context with vision and video support.

Higher quality

Model	Role
Qwen/Qwen3.5-397B-A17B-TEE	Large Qwen model for complex reasoning.
Qwen/Qwen3-235B-A22B-Thinking-2507-TEE	Thinking variant for multi-step reasoning.

Environment variables

Variable	Purpose
CHUTES_API_KEY	API key instead of configure-stored credentials
TOBY_PRETREAT_MODEL	Override the pretreatment model (e.g. deepseek-ai/DeepSeek-V3.2-TEE)
TOBY_PRETREAT_DELTA	Set to 0 to disable follow-up delta pretreatment

See also

• AI providers overview
• OpenRouter — another multi-model provider
• Personas