Serverless Inference
Send requests to Crusoe-hosted models over an OpenAI-compatible API or interact with them directly using the Intelligence Foundry's chat interface. Crusoe handles serving, scaling, and optimization—so you don't have to manage any GPU infrastructure.
1. Log in or create an account
Log in to the Crusoe Cloud Console or Create an account.
After you log in, switch to the Intelligence Foundry app in the bottom-left of the console.
2. Generate an API key
To create an API key via the console:
- From the console, click Admin in the bottom-left corner.
- Select Security > Intelligence API keys from the left navigation.
- Click Create.
- (Optional) Enter an alias for your key.
- (Optional) Enter an expiration date for your key.
- Copy the API key. Make sure that you save the key in a secure location before leaving the page.
For more information, see Manage API keys.
3. (Optional) Browse available models
You can use the OpenAI-API compatible endpoint at api.inference.crusoecloud.com to access the models below for Serverless Inference. You can also interact with all of the models using the Intelligence Foundry's chat interface. All Meta models provided by Crusoe are "Built with Llama".
For each model's pricing information, see pricing.
| MODEL | PROVIDER | TYPE | CONTEXT LENGTH | LICENSE | ACCEPTABLE USE POLICY |
|---|---|---|---|---|---|
| deepseek-ai/DeepSeek-V4.1-Flash | DeepSeek | instruct | 1M | MIT License | |
| google/gemma-4-31b-it | instruct | 262k | Apache License 2.0 | ||
| nvidia/Nemotron-3-Nano-Omni-Reasoning-30B-A3B | NVIDIA | instruct | 262k | NVIDIA Open Model Agreement | |
| nvidia/Nemotron-3-VoiceChat | NVIDIA | speech-to-speech | 131k | NVIDIA Software and Model Evaluation License | NVIDIA Acceptable Use Terms |
| nvidia/nemotron-3.5-lightning-30b-a3b | NVIDIA | instruct | 262k | NVIDIA Nemotron Open Model License | NVIDIA Acceptable Use Terms |
| nvidia/nemotron-3.5-super-vl-120b-a12b | NVIDIA | instruct | 1M | NVIDIA Nemotron Open Model License | NVIDIA Acceptable Use Terms |
| openai/gpt-oss-120b | OpenAI | instruct | 128k | Apache License 2.0 | Acceptable Use Policy |
| qwen/Qwen3.8-27B | Qwen | instruct | 256k | Apache License 2.0 | Apache License 2.0 |
| zai-org/GLM-5.3 | Z.ai | instruct | 1M | MIT License | |
| zai-org/GLM-5.3-Flash | Z.ai | instruct | 1M | MIT License |
Migrate from a deprecated model
If you were using a model that is no longer in the available models table above, you need to migrate to a supported replacement:
- Find your deprecated model in the table below and identify its recommended replacement.
- Update your API calls to reference the replacement model and test the replacement model in your development environment.
The following table lists each deprecated model, the date it was deprecated, and the recommended replacement:
| Deprecated model | Migration path | Deprecation date |
|---|---|---|
| DeepSeek V4 Pro | DeepSeek V4.1 Flash | November 2, 2026 |
| DeepSeek V4 Flash 0731 | DeepSeek V4.1 Flash | November 2, 2026 |
| NVIDIA Nemotron 3 Super 120B | NVIDIA Nemotron 3.5 Super VL | November 2, 2026 |
| NVIDIA Nemotron 3 Nano | NVIDIA Nemotron 3.5 Lightning 30B | November 2, 2026 |
| Moonshot Kimi K2.6 | DeepSeek V4.1 Flash or Z.ai GLM-5.3 | October 2, 2026 |
| Z.ai GLM-5.1 | Z.ai GLM-5.3 | September 12, 2026 |
| Z.ai GLM-5.2 | Z.ai GLM-5.3 | September 12, 2026 |
| DeepSeek V3 | DeepSeek V4.1 Flash | September 12, 2026 |
| Qwen3 235B A22B | DeepSeek V4.1 Flash or Qwen 3.8 27B | September 12, 2026 |
| Llama 3.3 70B | DeepSeek V4.1 Flash | September 12, 2026 |
For the current list of supported models, see Available models.
4. Send a request
Use the OpenAI-compatible endpoint at api.inference.crusoecloud.com. The example below queries deepseek-ai/DeepSeek-V4.1-Flash:
Need higher rate limits or reserved capacity?
When interacting with a Serverless endpoint, you might receive a 429 Too Many Requests response due to rate limits. If you need to exceed the default rate limits, you can:
- Contact us for a rate limit increase. This is recommended if you expect your initial launch traffic to exceed the default limits.
- If you need reserved inference capacity, use Self-Serve Deployments.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.getenv("CRUSOE_API_KEY"),
base_url="https://api.inference.crusoecloud.com/v1",
)
completion = client.chat.completions.create(
model="deepseek-ai/DeepSeek-V4.1-Flash",
messages=[
{"role": "system", "content": "You are a helpful, concise assistant."},
{"role": "user", "content": "Who is Robinson Crusoe?"},
],
)
print(completion.choices[0].message.content)
Additional resources: