Llama
Meta's open-weights family of foundation models
Overall score
About
Llama is Meta's family of open-weights large language models. Models can be downloaded and run locally, hosted on managed inference providers, or fine-tuned on custom data. Used as the default open-source baseline by builders who want to avoid lock-in to closed-source frontier APIs.
Best for: Builders and engineering teams that want a free, open-weights frontier model they can self-host, fine-tune, or run on any inference provider without vendor lock-in.
Pricing
Open Weights
- Monthly
- n/a
- Annual /mo
- Free
- Billing
- flat
- Notes
- All Llama model weights;Custom commercial licence (up to 700M MAU);Self-host on your own GPUs · Inference costs are paid to whichever provider hosts the model (Together AI, Groq, AWS, etc.) or your own infrastructure.
| Tier | Monthly | Annual /mo | Billing | Notes |
|---|---|---|---|---|
| Open Weights | n/a | Free | flat | All Llama model weights;Custom commercial licence (up to 700M MAU);Self-host on your own GPUs · Inference costs are paid to whichever provider hosts the model (Together AI, Groq, AWS, etc.) or your own infrastructure. |
Key features
- Open weights (custom commercial licence)
- Multiple model sizes (1B to 405B)
- Multimodal variants
- Fine-tuneable on consumer hardware at smaller sizes
- Extensive third-party hosting
Integrations
Trust & compliance
- Stage range
- n/a
- Founded
- 2023
- Status
- active
- SOC 2
- n/a
- GDPR
- n/a
- Data residency
- customer_choice
- External rating
- n/a
- Last verified
- Jun 2026
Reviews
Be the first to share your experience.
Related tools in Agent infrastructure
Pairs well with
People who've discussed Llama
See all people →Curated mentions from podcasts, posts, and public stacks. Editorial coverage; not endorsements.
Ali Ghodsi
Co-founder & CEO
Ali Ghodsi platforms Meta's Llama family as one of the four supported model choices on Databricks. The open-weights option is part of his broader thesis that customers should be able to combine multiple model providers (OpenAI, Anthropic, Google, Meta) with their own data rather than locking into a single closed-API vendor.
Clem Delangue
Co-founder & CEO
Clem Delangue treats Meta's Llama family as one of the canonical open-weights reference models, alongside DeepSeek R1. In his Ultra-Scale Playbook X post he uses Llama as a benchmark example for what frontier-model training should look like at every tech company. The Llama family is one of the most-downloaded model families on the Hugging Face hub he runs.
Yann LeCun
Chief AI Scientist
LeCun is Chief AI Scientist at Meta and Llama is Meta's open-source family of foundation models. He's championed Llama in nearly every public talk and X thread, framing open-source weights as the path to durable progress on AI rather than the closed-frontier-model approach taken by OpenAI and Anthropic. Every Llama release gets his commentary as the public-research-vs-product framing point.