Shared endpoints provide instant access to popular models for development and testing purposes. They’re designed for:
  • Quick experimentation and prototyping
  • Development and testing environments
  • Learning and evaluation of models
  • Proof of concept development
Note: Shared endpoints are not intended for production use. For production workloads, we strongly recommend using private deployments.

​
Quick Start

First, install the Predibase Python SDK: 
pip install -U predibase
Then you can start experimenting with shared endpoints using just a few lines of code: 
from predibase import Predibase

pb = Predibase(api_token="<PREDIBASE_API_TOKEN>")

# Use a shared endpoint for testing
client = pb.deployments.client("qwen3-8b")

# Generate text
response = client.generate("What is machine learning?")
print(response.generated_text)

​
Available Models

Predibase offers several popular models as shared endpoints for testing and development. See our supported models for the complete list.

​
Using Shared Endpoints

​
With Python SDK

Here’s a detailed example showing both basic text generation and streaming responses for development: 
from predibase import Predibase

pb = Predibase(api_token="<PREDIBASE_API_TOKEN>")
client = pb.deployments.client("qwen3-8b")

# Basic generation with customizable parameters
response = client.generate(
    "Explain quantum computing in simple terms.",
    max_new_tokens=100
)
print(response.generated_text)

# Stream responses for real-time output
for response in client.generate_stream(
    "Write a story about a robot learning to paint.",
    max_new_tokens=200
):
    print(response.token.text, end="", flush=True)

​
With REST API

For testing language-agnostic integration, you can use our REST API: 
# Set your credentials
export PREDIBASE_API_TOKEN="<YOUR TOKEN>"
export PREDIBASE_TENANT_ID="<YOUR TENANT ID>"
export PREDIBASE_MODEL="qwen3-8b"

# Generate text
curl -d '{
    "inputs": "Explain quantum computing in simple terms.",
    "parameters": {
        "max_new_tokens": 100
    }
}' \
    -H "Content-Type: application/json" \
    -X POST https://serving.app.predibase.com/$PREDIBASE_TENANT_ID/deployments/v2/llms/$PREDIBASE_MODEL/generate \
    -H "Authorization: Bearer $PREDIBASE_API_TOKEN"

​
Testing Custom Models

You can test your fine-tuned adapters on shared endpoints during development: 
# Test your custom adapter with a shared endpoint
response = client.generate(
    "Summarize this article.",
    adapter_id="my-summarizer/1",  # Your adapter ID
    max_new_tokens=100
)
For more information about using fine-tuned models, see:

​
Rate Limits

Shared endpoints are subject to rate limits. Rate limits are restrictions that our API enforces on how often users can access our services within a given time period and can be identified via HTTP 429 error codes.

​
Rate Limits by Tier

TierRate LimitDailyMonthly
Free1 request / sec1 million tokens / day10 million tokens / day
Enterprise SaaS100 requests / sec1 million tokens / day10 million tokens / day
Enterprise VPCDoes not applyDoes not applyDoes not apply

​
Rate Limit Headers

When making API requests, you’ll receive the following headers that help you monitor your rate limit status:
HeaderDescription
x-envoy-ratelimitedWhether the rate limit has been reached
x-ratelimit-limitThe max number of requests until the rate limit is reached
x-ratelimit-remainingThe remaining number of requests until the rate limit is reached
x-ratelimit-resetAmount of time (seconds) until you can query again
For production use cases, you should use private deployments which do not have any rate limits.

​
Moving to Production

When you’re ready to move your application to production:
  1. Set up a private deployment for production-grade reliability
  2. Configure auto-scaling to handle your workload
  3. Take advantage of dedicated resources and SLAs

​
Next Steps