Crusoe

Overview

Property	Details
Description	Crusoe Cloud provides GPU-accelerated inference for open-source large language models, optimized for performance and cost efficiency.
Provider Route on LiteLLM	crusoe/
Link to Provider Doc	Crusoe Serverless Inference Documentation ↗
Base URL	https://api.inference.crusoecloud.com/v1 (older LiteLLM versions call a different host by default, see Custom API Base)
Supported Operations	/chat/completions

We support ALL Crusoe models, just set crusoe/ as a prefix when sending completion requests

Available Models

Model	Provider	Context Window
crusoe/deepseek-ai/Deepseek-V4-Flash	DeepSeek	1,048,576 tokens
crusoe/deepseek-ai/DeepSeek-V4.1-Flash	DeepSeek	1,048,576 tokens
crusoe/deepseek-ai/DeepSeek-V4-Pro	DeepSeek	1,048,576 tokens
crusoe/google/gemma-4-31b-it	Google	262,144 tokens
crusoe/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B	NVIDIA	262,144 tokens
crusoe/nvidia/NVIDIA-Nemotron-3.5-Super-VL	NVIDIA	262,144 tokens
crusoe/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B	NVIDIA	262,144 tokens
crusoe/nvidia/Nemotron-3-Nano-Omni-Reasoning-30B-A3B	NVIDIA	262,144 tokens
crusoe/nvidia/Nemotron-3.5-Lightning-30B-A3B	NVIDIA	262,144 tokens
crusoe/openai/gpt-oss-120b	OpenAI	131,072 tokens
crusoe/Qwen/Qwen3.8-27B	Qwen	262,144 tokens
crusoe/yutori/n2	Yutori	262,144 tokens
crusoe/zai-org/GLM-5.3	Z.ai	1,048,576 tokens
crusoe/zai-org/GLM-5.3-Flash	Z.ai	1,048,576 tokens

Crusoe retires models often. To list the model ids your key can call, send GET https://api.inference.crusoecloud.com/v1/models with your Crusoe API key, and put crusoe/ in front of each id. Copy ids from that endpoint rather than from Crusoe's docs pages, which spell some of them differently.

LiteLLM's cost map does not have current prices for these models, so their spend logs as $0 or at an outdated price. To track spend, set input_cost_per_token and output_cost_per_token for each model you use from Crusoe's pricing, in the model's litellm_params on the proxy or as completion() arguments in the SDK.

Required Variables

Environment Variables

os.environ["CRUSOE_API_KEY"] = ""  # your Crusoe API key

Usage - LiteLLM Python SDK

Non-streaming

Crusoe Non-streaming Completion

import os
import litellm
from litellm import completion

os.environ["CRUSOE_API_KEY"] = ""  # your Crusoe API key

messages = [{"content": "Hello, how are you?", "role": "user"}]

# Crusoe call
response = completion(
    model="crusoe/deepseek-ai/Deepseek-V4-Flash",
    messages=messages
)

print(response)

Streaming

Crusoe Streaming Completion

import os
import litellm
from litellm import completion

os.environ["CRUSOE_API_KEY"] = ""  # your Crusoe API key

messages = [{"content": "Write a short story about AI", "role": "user"}]

# Crusoe call with streaming
response = completion(
    model="crusoe/deepseek-ai/Deepseek-V4-Flash",
    messages=messages,
    stream=True
)

for chunk in response:
    print(chunk)

Function Calling

Crusoe Function Calling

import os
import litellm
from litellm import completion

os.environ["CRUSOE_API_KEY"] = ""  # your Crusoe API key

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get the current weather in a location",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {
                    "type": "string",
                    "description": "The city and state, e.g. San Francisco, CA"
                }
            },
            "required": ["location"]
        }
    }
}]

messages = [{"role": "user", "content": "What's the weather in Boston?"}]

response = completion(
    model="crusoe/deepseek-ai/Deepseek-V4-Flash",
    messages=messages,
    tools=tools,
    tool_choice="auto"
)

print(response)

Usage - LiteLLM Proxy Server

config.yaml

model_list:
  - model_name: deepseek-v4-flash
    litellm_params:
      model: crusoe/deepseek-ai/Deepseek-V4-Flash
      api_key: os.environ/CRUSOE_API_KEY
  - model_name: deepseek-v4-pro
    litellm_params:
      model: crusoe/deepseek-ai/DeepSeek-V4-Pro
      api_key: os.environ/CRUSOE_API_KEY
  - model_name: gpt-oss-120b
    litellm_params:
      model: crusoe/openai/gpt-oss-120b
      api_key: os.environ/CRUSOE_API_KEY
  - model_name: qwen3.8-27b
    litellm_params:
      model: crusoe/Qwen/Qwen3.8-27B
      api_key: os.environ/CRUSOE_API_KEY
  - model_name: glm-5.3
    litellm_params:
      model: crusoe/zai-org/GLM-5.3
      api_key: os.environ/CRUSOE_API_KEY

Custom API Base

Option 1: Environment variable

Older LiteLLM versions, including v1.106 and earlier, send Crusoe calls to https://managed-inference-api-proxy.crusoecloud.com/v1 by default. That host's firewall answers some prompts with an HTTP 403 "Sorry, you have been blocked" page, for example prompts that contain file paths like ../../etc/passwd or SQL injection strings like ' OR 1=1 --. To see which host your version calls, send a request that succeeds and read the x-litellm-model-api-base response header on the proxy, or response._hidden_params["api_base"] in the SDK. A blocked request does not show the host. If it is the old host, set CRUSOE_API_BASE to https://api.inference.crusoecloud.com/v1 as below. If your deployment restricts outbound traffic, allow the host your version calls.

Custom API Base via env var

import os
from litellm import completion

os.environ["CRUSOE_API_BASE"] = "https://api.inference.crusoecloud.com/v1"
os.environ["CRUSOE_API_KEY"] = ""  # your API key

response = completion(
    model="crusoe/deepseek-ai/Deepseek-V4-Flash",
    messages=[{"content": "Hello!", "role": "user"}],
)

Option 2: Pass directly

Custom API Base via parameter

from litellm import completion

response = completion(
    model="crusoe/deepseek-ai/Deepseek-V4-Flash",
    messages=[{"content": "Hello!", "role": "user"}],
    api_base="https://api.inference.crusoecloud.com/v1",
    api_key="your-api-key",
)

Supported OpenAI Parameters

• temperature
• max_tokens
• max_completion_tokens
• top_p
• frequency_penalty
• presence_penalty
• stop
• n
• stream
• tools
• tool_choice
• response_format
• seed
• user
• logit_bias
• logprobs
• top_logprobs