Llama logo

Llama

Verified Verified
by Meta

Meta's open-weight model family — the default foundation for self-hosted AI.

Free Plan Available Open Source API Available
4.6 Rating
1.2K reviews
22.2K+ visits
Visit Website
No preview available

About Llama

Meta’s open-weight LLM family for research and production.
  • Conversational Chat AI Capabilities
  • Text Generation Content & Media
  • Public API Developer & Automation
  • Export & Download Collaboration
  • Data-Retention Controls Security & Compliance

Key Features

AI Capabilities

  • Conversational Chat

    Multi-turn natural-language conversation with follow-up context.

  • Reasoning Mode

    A deliberate "thinking" mode that works through multi-step problems before answering.

  • Multilingual Support

    Understands and replies in dozens of languages, including translation between them.

  • Long-Context Understanding

    Handles very long documents and conversations without losing earlier context.

Content & Media

  • Text Generation

    Drafts, rewrites and edits long-form and short-form copy.

  • Image Understanding

    Vision models

    Reads photos, screenshots, diagrams and handwriting and answers questions about them.

Developer & Automation

  • Public API

    Via providers

    A documented API for building the same capability into your own product.

  • Code Generation & Debugging

    Writes, explains, reviews and fixes code across common languages.

  • Custom Assistants

    Fine-tuning

    Build and share a configured assistant with its own instructions and knowledge.

  • Task Automation

    Schedules recurring jobs or runs multi-step tasks without supervision.

  • Third-Party Integrations

    Huge ecosystem

    Connects to other SaaS products through apps, connectors or an integration platform.

Collaboration

  • Export & Download

    Open weights

    Export chats, documents or data out of the product in an open format.

Llama Pricing

Prices are set by the vendor and may change.

Popular

Open weights

$0

What's included

  • Free under the Llama Community Licence
  • Run on your own hardware
  • Fine-tune without restriction
  • No per-token cost
Download

Cloud providers

Per token usage

What's included

  • Hosted on AWS, Azure, Google Cloud and others
  • Pay only for what you use
  • No infrastructure to manage
  • Competitive per-token rates
Compare providers

Enterprise licence

Custom

What's included

  • Required above a large user threshold
  • Negotiated commercial terms
  • For the largest deployments
  • Direct engagement with Meta
Contact Meta
View full pricing details

What Users Say

4.6/5
Based on 1.2K reviews

No written reviews yet

Be the first to share your experience with Llama.

Write the first review
Share your experience
Log in to write a review for Llama.
Log in to review

Real-World Use Cases

  • Self-hosting
  • Fine-tuning
  • On-premise deployment
  • Research
  • Cost control
  • Edge devices

What is Llama?

Llama is Meta's family of open-weight large language models. Meta publishes the weights under a permissive community licence, so anyone can download, run, fine-tune and deploy them without paying per token or sending data to a third party.

That decision reshaped the industry. Llama is now the default base model for self-hosted AI, and the ecosystem around it — Ollama, llama.cpp, vLLM, thousands of community fine-tunes — is larger than any other open family's.

Sizes range from small models that run on a laptop to large ones needing serious GPU clusters, so the same family covers edge devices and datacentre workloads.

Benefits of Llama

Why it became the default:

  • Open weights. Download, inspect, modify, deploy — no permission needed.
  • No per-token cost once you are running your own inference.
  • Complete data control. Nothing leaves your infrastructure.
  • A size for every constraint, from laptop to cluster.
  • The largest open ecosystem of tooling and fine-tunes.
  • Available on every major cloud if you would rather not self-host.

Who Should Use Llama?

Llama is infrastructure, not a consumer product.

  • Engineering teams building AI features where API cost or latency matters.
  • Regulated industries that cannot send data to a third-party API.
  • Researchers needing weights they can study and modify.
  • Startups fine-tuning a base model for a narrow domain.
  • Edge and offline deployments.

If you just want to chat with an AI, use a hosted assistant — Llama requires engineering to become a product.

Pros & Cons

6Pros 4Cons

Pros

6
  • Open weights — download, modify and deploy with no vendor lock-in.
  • Zero per-token cost once you run your own inference.
  • Complete data control; nothing leaves your infrastructure.
  • Model sizes span laptop-scale to datacentre-scale.
  • The largest tooling and fine-tune ecosystem of any open family.
  • Available on every major cloud if you prefer a hosted route.

Cons

4
  • Not a finished product — it takes engineering to deploy.
  • The community licence has conditions, including a large-scale user threshold.
  • Self-hosting the biggest models requires expensive GPU hardware.
  • Top-end quality still trails the leading closed frontier models.

What's New

  1. Expanded context windows

    Feature

    Longer context support across the model family.

  2. Vision-capable models

    Feature

    Image understanding added to the open-weight releases.

  3. Licence clarification

    News

    Updated community licence terms and usage threshold.

  4. Small on-device models

    Feature

    Compact variants targeted at phones and edge hardware.

Frequently Asked Questions

Is Llama really free?
The weights are free to download and use under the Llama Community Licence. A separate commercial licence is required only above a very large monthly-user threshold, which affects almost no one.

Is Llama open source?
It is "open weight" rather than strictly open source. The weights are published and freely usable, but the licence includes conditions that the OSI definition of open source does not allow.

What hardware do I need?
The smallest models run on a laptop with 8GB of RAM. Mid-size models want a 24GB GPU. The largest need multi-GPU servers — or use a hosted provider instead.

How do I run Llama locally?
Ollama is the easiest route — one command downloads and runs a model. llama.cpp and LM Studio are popular alternatives, and vLLM is the standard for production serving.

Llama or a hosted API?
Self-host when data control, cost at scale or offline operation matter. Use a hosted API when you want the best quality with no infrastructure work.

Can I fine-tune Llama on my own data?
Yes, and this is one of its main attractions. LoRA fine-tuning is achievable on modest hardware; full fine-tuning needs considerably more.

Tool Information

  • Category AI Models & Platforms
  • PricingOpen Source
  • Free PlanYes
  • API AvailableYes
  • Mobile AppNo
  • Language Support16+ Languages
  • Starting Price$0
  • DeveloperMeta
  • Founded2023
  • Websitewww.llama.com
  • StatusActive
  • Last UpdatedAug 15, 2026
Follow

Best For

Self-hostingFine-tuningOn-premise deploymentResearchCost controlEdge devices

Platforms

Web Windows macOS Linux API

Tags

LLMFree PlanAPI Available

Languages

Arabic Chinese Dutch English French German Hindi Indonesian Italian Japanese Korean Polish Portuguese Russian Spanish Turkish

Share Llama

Love this tool? Share it with your friends!

Own this listing?

Claim Llama to manage the profile, publish updates and see your analytics.

Claim This Listing

Verification takes a couple of minutes.

Ready to experience the power of Llama?

Join the users who trust Llama for their daily tasks.

Try Llama Now