Hosted open-source model catalog
Access models from families including Qwen, DeepSeek, GLM, Kimi, Gemma, Mistral, and others through Chutes-hosted inference endpoints.
Chutes
Freemium
Chutes is a decentralized, serverless AI inference platform for deploying, running, and scaling open-source models. Developers can access hosted models through an API or deploy private custom workloads on confidential GPU capacity.
什么是 Chutes？
Chutes is a decentralized, serverless inference platform for open-source AI models. Developers can use hosted models through an API, run models from the platform catalog, or deploy custom code as a private chute on confidential GPU capacity.
The platform is designed for AI-powered products and workloads ranging from language inference to image, video, speech, music, embeddings, and content moderation. Its documented developer workflow includes SDK and CLI tools, API-key authentication, container builds, GPU selection, endpoint exposure, and deployment.
Public inference is priced by token, while private deployments are billed for GPU runtime plus a one-time deployment fee. Chutes also offers monthly plans and an enterprise contact path for organizations needing volume discounts, custom rate limits, or dedicated support.
Chutes 能做什么？
API and SDK access
Send inference requests through the API, including streaming chat completions, and use the Chutes SDK and CLI to create and operate deployments.
Private custom deployments
Build a container, define the required GPU class, VRAM, and count, expose endpoints through the chute definition, and deploy the workload privately.
Confidential GPU compute
Private Chutes run on isolated confidential GPU capacity. The pricing page states that featured public models also run on confidential TEE compute.
Multiple AI modalities
The platform lists support for language models as well as image, video, speech, music, embedding, and content-moderation workloads.
Usage-based and subscription billing
Public inference can be paid for by input and output tokens, while private deployments use GPU runtime billing; Plus, Pro, and enterprise options are also available.
使用场景
“Production LLM inference”
Use always-on hosted language models behind an API for AI-powered products, chat features, reasoning, coding, or agent workloads.
“Custom model serving”
Package your own model code or a custom fine-tune into a container and deploy it as a private chute with a selected GPU configuration.
“Multimodal generation and processing”
Use catalog models for image, video, speech, music, embedding, or content-moderation tasks instead of limiting an application to text generation.
“Agent and coding workflows”
Route model calls from OpenClaw, Hermes, Claude Code, or MCP-compatible clients to Chutes-backed models for terminal, agent, and development workflows.
“Workflow automation and hosted chat”
Connect Chutes inference to n8n automations or use the Open WebUI integration for a chat interface with Chutes-specific deployment support.
常见问题
How can developers access Chutes models?
They can call the documented inference API with an API key, use the Chutes SDK and CLI, or connect supported developer and automation tools such as OpenClaw, Hermes, n8n, Open WebUI, MCP clients, and Claude Code.
Can I deploy my own model or code?
Yes. Chutes documents a private deployment workflow in which you build a container, define a GPU NodeSelector, expose endpoints in the chute definition, and run the deploy command.
How is Chutes inference billed?
Public inference uses per-token pricing based on input and output tokens. Private Chutes are billed by GPU runtime and include a one-time deployment fee. Monthly Plus and Pro plans and an enterprise sales option are also available.
What kinds of models and workloads are supported?
The platform focuses on open-source AI models and lists language, image, video, speech, music, embedding, and content-moderation workloads. Specific availability depends on the model catalog and deployment configuration.
Are private deployments always running?
The pricing page states that idle private instances shut down automatically, so private deployment charges apply while the instance is running.
快速信息
- Category
- Serverless AI inference and compute
- Primary users
- AI developers and teams building model-powered products
- Model focus
- Open-source models
- Access methods
- API, SDK, CLI, hosted platform, and integrations
- Deployment options
- Hosted public inference or private custom Chutes
- Billing models
- Per-token inference, GPU runtime, monthly plans, and enterprise billing
Chutes 替代品
SGLang
www.sglang.io
SGLang is an open-source serving framework for large language and multimodal models. It helps developers deploy model servers and query them through OpenAI-compatible endpoints across supported GPU, CPU, TPU, NPU, and XPU platforms.
Replicate
replicate.com
Run, fine-tune, and deploy machine-learning models through a cloud API.
Cerebrium
cerebrium.ai
Serverless CPU and GPU infrastructure for teams deploying real-time AI workloads.
UbiOps
ubiops.com
Model serving and orchestration for AI teams deploying workloads across cloud and private infrastructure.