Chutes
Decentralized serverless GPU platform for open-source model inference and deployment.
About
Chutes is a decentralized serverless AI compute platform built on Bittensor Subnet 64. It exposes an OpenAI-compatible inference endpoint over a catalog of open-source models including DeepSeek, Qwen, MiniMax, Kimi, and GLM, letting developers point existing OpenAI SDK code at it with only a base URL change.
Developers can also deploy their own chutes, which are complete AI applications packaged with FastAPI-style endpoints, Docker images, hardware selection, and auto-scaling, billed pay-per-use without managing Kubernetes. As of 2026 the entire hosted LLM catalog runs in trusted execution environments for confidential compute, and subscription tiers offer usage allowances at a multiple of pay-as-you-go value. In September 2026 Chutes demonstrated Parallax, a model trained on 240 distributed consumer GPUs, underscoring its decentralized training story.
What it can do
- OpenAI-compatible serverless inference for open models
- Deployable chutes with auto-scaling and hardware selection
- Confidential compute via trusted execution environments
- Subscription tiers with capped usage allowances
- Image, video, speech, music, and embedding workloads
- Vercel AI SDK and MCP integrations
Setup
Create an account at chutes.ai, generate an API key, and point any OpenAI-compatible client at the Chutes endpoint.
Prerequisites
A Chutes account; API key for inference or CLI for deploying custom chutes.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 91; Trending score 76/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.