Serverless AI compute from Chutes, tuned for any scale.
Chutes runs open-source models on a decentralized GPU network, giving teams a single API to deploy, scale and serve them in production, from first prototype to trillions of tokens a month.
OpenRouter
Kilo
FetchAI
Phala
Pax Historia
Fresh open-source LLMs, live on day one.
Our engineers track every major release and bring new open-weight models online within minutes. When a lab ships something big, Chutes is usually the first place you can call it, so you can build on what comes next today.
Browse the LLMs
Quick call
Python
import os
import requests
resp = requests.post(
f"{os.environ['API_BASE']}/v1/chat/completions",
headers={"Authorization": f"Bearer {os.environ['API_KEY']}"},
json={
"model": "moonshotai/Kimi-K2.6-TEE",
"messages": [{"role": "user", "content": "Draft a release checklist."}],
"stream": True,
},
stream=True,
)
for line in resp.iter_lines():
if line.startswith(b"data: "):
print(line.decode()[6:])A Chute for every modality
Language models are only the start. Chutes also hosts image, video, speech and music models, covering every open-source modality, kept warm and ready to scale.
Sam3
sam3
Locate and segment objects inside any picture
Z image turbo
z-image-turbo
Quick text-to-image renders in a few steps
Qwen Image Edit 2511
Qwen-Image-Edit-2511
Change pictures using plain-language prompts
Imageclassic
imageclassic
FLUX.1-schnell and three SDXL checkpoints, packed into one of our chutes
Built for hyperscale AI products
Fast inference for leading open-source models, plus short-lived jobs, batch pipelines and plenty more. Bring your code to Chutes and the platform takes care of everything else.
Launch in minutes
Made for AI builders
Flexible by design, quick in practice
Open, decentralized infrastructure
Platform capabilities
Model inference
Always-warm models that stay steady under load.
TEE secure compute
Private, isolated hardware for sensitive workloads.
Consumer apps
Chutes Chat for everyday users, and fictio for character roleplay, both running on this network.
Two ways to build on Chutes
Start with a hosted model in minutes, or package your own and let the network handle scaling.
Hosted models
Use hosted chutes through one OpenAI-style endpoint
- Chat, completion and embedding routes
- Streaming responses built in
- A catalog that keeps growing
Bring your own
Wrap any open model or script and deploy it
- Ship Python code or a container
- Stay warm or scale down when idle
- Optional TEE isolation
Unsure which route suits your workload? Our team can help you map it out.
Ask the teamLeading models, online and ready to grow.
Bring your own code or pick a hosted model. On Chutes you skip cold starts and capacity worries, and every listed model stays reachable.
Hy4
tencent/Hy4
Text and vision model behind an OpenAI-style API
Qwen3.8 27B TEE pareton
Qwen/Qwen3.8-27B-TEE-pareton
Description coming soon.
GLM 5.2 Fast
zai-org/GLM-5.2-Fast
Low-latency language model, OpenAI-style routes
Qwen3.8 2.4T A95B
Qwen/Qwen3.8-2.4T-A95B
Large mixture-of-experts model for text and vision
DeepSeek V4.1 Flash 1017
deepseek-ai/DeepSeek-V4.1-Flash-1017
Speedy reasoning model with OpenAI-style routes
GLM 5.3
zai-org/GLM-5.3
General-purpose model for chat, code and agents
Kimi K2.7 Code
moonshotai/Kimi-K2.7-Code
Coding-focused model for long agent sessions
Biomni R0 32B Preview FP8 Dynamic
nordforge/Biomni-R0-32B-Preview-FP8-Dynamic
Description coming soon.
DeepSeek V4 Pro 0813
deepseek-ai/DeepSeek-V4-Pro-0813
Flagship reasoning model for demanding tasks
Qwen3.8 Max 0902
Qwen/Qwen3.8-Max-0902
Top-tier Qwen release for text and images