Packages and Models

Start With the Idea, Not the Setup

Get to work with Anaconda’s vetted packages and curated AI models. Use them in your favorite tools or in Kilo Desktop to explore, experiment, and build with AI.
Get Started for Free

AI Projects Shouldn't Start With a Scavenger Hunt

Before you start to research data or write a line of real code, you hunt for packages, compare models from public sources, fix dependency conflicts, and wait for someone to approve it all. Your AI project has to wait.

Setup Slows You Down

New projects start with the same chores: find the packages, pick a model, untangle dependency conflicts. The time you meant for building goes to assembly.

You Can't Tell What's Safe to Pull

A public repo won’t tell you who really built a model, what its license allows, or which known CVEs came along with a package. Guess wrong and it’s your problem.

Reviews Show Up After You've Built

You’re ready to move forward, then a package or model needs another review. Late approvals interrupt your momentum and can send you back to replace what you’ve already built.

Builders Want Speed. Enterprises Want Proof. You Need Both.

For more than a decade, Anaconda has helped developers and data scientists work with open source. Today, 50M+ users and 95% of the Fortune 500 build on Anaconda.

One Trusted Source for the Packages and Models You Build On

Get Python packages built from source, tested, and signed by Anaconda, alongside curated AI models. With dependency, license, and vulnerability details in one place, you spend less time investigating what you’re installing and more time using it.

Start With Packages Built and Tested by Anaconda

Pull from 19K+ packages built from the source, tested for quality, and signed by Anaconda. Dependency, license, and vulnerability details are available when you need them, so you spend less time researching packages and more time building.

  • Packages built from source and tested by Anaconda
  • Dependency, license, and vulnerability details
  • Installation through conda in your existing workflow
  • Custom channels and mirroring
Learn More →

Choose the Right Model With the Facts Up Front

Browse 77+ curated open-source models. Compare benchmark scores, check disk and RAM needs for each quantization, and review the license and AI Bill of Materials, then start a model server or open it in Kilo Desktop.

  • Side-by-side benchmark comparison
  • Quantizations sized to your hardware
  • Downloadable AIBOM (CycloneDX)
  • Trusted Publisher badges
Start with Kilo Desktop →

Let Policy Follow the Work, Not a Review Queue

Your security team can optionally set policies for which packages and models are available, helping you choose options that meet your organization’s requirements. Spend less time going back and forth on approvals and more time building.

  • Packages and models filtered by your organization’s policies
  • Automatic re-checks on mirror updates
  • Available choices in the tools you already use
  • Policy reports and change history
Learn More →

Get From Idea to Build in Three Steps

You work with your same tools. We set up the trusted source underneath.

Step 1

Download Kilo Desktop for AI development with notebooks and agentic tools, or access Anaconda packages directly from your existing environment.

Step 2

Explore vetted Python packages and curated AI models. Find what you need for your next analysis, experiment, or application.

Step 3

Install packages and pull models into your workflow. Keep working in VS Code, JetBrains, Jupyter, the CLI, or Kilo Desktop.
Start Here

Spend Your Time on the Idea, Not the Inputs

Builders move faster, security sees what’s in use, and security teams see what’s in use and stops acting as the gate.

Spend More Time Bringing Ideas to Life

Find your building blocks, make informed choices, and keep building in the tools you know.

Start With Less Setup

Find vetted packages and curated models in one place, ready to use in your development workflow.

Choose With Confidence

Compare model benchmarks and hardware requirements, and check package dependencies and licenses with details close at hand.

Keep Your Momentum

Your security team’s policies help clarify which packages and models you can use, reducing approval back-and-forth as you build.

Get Your Next Idea Building

Find trusted packages and curated AI models for your next analysis, experiment, or application. Start with the tools you already use.
Get Started for Free

Frequently Asked Questions

How does Anaconda vet the packages in its repository?

Packages in Anaconda’s repository are built and tested by Anaconda before release, then cryptographically signed so you can confirm they haven’t been tampered with since leaving our pipeline. Anaconda syncs with the National Vulnerability Database (NVD) to track CVEs across packages, and provides Software Bills of Materials (SBOMs) for popular packages in SPDX format.

Packages docs

The model catalog is a curated set of security-reviewed, open-source models. Each model lists its publisher, license, intended use, benchmark scores, and the disk and RAM needed for each quantization. Models with a Trusted Publisher badge come from organizations Anaconda has verified for legal standing, security practices, documentation standards, and community presence.

Model catalog docs | Kilo Desktop

You can download each model’s AIBOM as a CycloneDX JSON file. It can include model metadata and license terms, details for each quantization, SHA-256 checksums to verify files, benchmark scores, documented limitations and bias risks, and required software dependencies. Contents vary with what the publisher has disclosed and which benchmarks have been run.

Model Governance

Yes. Channel policies restrict packages by CVE score, CVE status, license, platform, package age, or conda spec, and reapply automatically as channels update. Model access policies allow or block models by name, publisher, license, purpose, tags, country of origin, parameter count, or expected VRAM. Blocked models are filtered out before they reach users. Policy filters are available on Business and Custom plans.

Policy docs

Yes. Developers and data scientists install packages with conda in VS Code, JetBrains, JupyterLab, PyCharm, Spyder, or the CLI. Models can run on a model server that exposes API endpoints for your applications, or open in Kilo Desktop to chat with and run locally.

IDE integration guides | Kilo Desktop