prime

What it is and what it does

Prime is the official CLI and Python SDK for Prime Intellect, a cloud platform offering GPU compute, isolated sandbox environments, and hosted training infrastructure. It lets you deploy GPU pods (H100, A100, etc.), create and manage remote sandboxes for code execution, launch training runs, browse pre-configured development environments, and manage team resources—all from the command line or Python code.

The package bundles a rich terminal UI (via textual and textual-plot), configuration management, SSH pod access, and both synchronous and asynchronous Python SDKs. It depends on httpx for HTTP requests, pydantic for data validation, and several Prime-specific packages (prime-sandboxes, prime-evals, prime-tunnel, verifiers) that appear to be separate but required runtime dependencies.

Use it for:

• Deploy and manage GPU pods for distributed machine learning training without managing infrastructure.
• Create isolated sandbox environments to run untrusted or experimental code safely.
• Launch and monitor hosted training runs with configurable models and resource allocation.
• Browse and install pre-configured development environments from the Prime Intellect community hub.
• Programmatically provision compute resources and execute remote commands via the Python SDK.
• Switch between personal and team contexts to manage shared resources and permissions.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

CLI and Python SDK for managing GPU compute, sandboxes, training runs, and environments on Prime Intellect's cloud platform.

Yes, if you use Prime Intellect's platform. The package is actively maintained, has low install friction, carries no license restrictions, and provides both CLI and SDK access to the service. Install only if you have a Prime Intellect account and API key; the CLI and SDK are tightly coupled to that platform and will not function standalone.

Install

prime on PyPI

pip install prime

uv add prime

poetry add prime

Installing prime

Before you install

Low install friction with a pure-Python wheel. Active maintenance with a release one day old and last commit on 2026-08-14. Requires Python 3.11–3.13; 16 runtime dependencies including httpx, pydantic, and textual for CLI rendering.

License in practice

MIT license (permissive) imposes no restrictions on use, modification, or distribution.

Quickstart

# Install
pip install prime

# Authenticate
export PRIME_API_KEY="your-api-key-here"

# CLI: list available GPUs
prime availability list

# Python SDK: create and manage a sandbox
from prime_sandboxes import APIClient, SandboxClient, CreateSandboxRequest
client = APIClient(api_key="your-api-key")
sandbox_client = SandboxClient(client)
sandbox = sandbox_client.create(CreateSandboxRequest(
    name="my-sandbox",
    docker_image="python:3.11-slim",
))

Requires a Prime Intellect account and valid API key (obtain from https://app.primeintellect.ai). Python 3.11–3.13 required.

Verify before relying

• Whether prime-sandboxes, prime-evals, prime-tunnel, and verifiers are separate packages or bundled; their availability and stability.
• Network connectivity and API key requirements for any CLI or SDK operation.
• Whether the CLI can function offline or requires active Prime Intellect account.
• GPU availability and pricing model on the Prime Intellect platform.

Package facts

Evidence: prime-0.6.22-py3-none-any.whl

Keywords: cli, cloud, compute, gpu

Tags

More Python Modules packages