Meta
✓ VerifiedCreator of the open-weights Llama model family.
At a glance
Meta AI
Meta Platforms (formerly Facebook, Inc.) operates the largest open-weights AI program in the industry through its Fundamental AI Research (FAIR) division, founded in 2013 and led by Turing Award winner Yann LeCun.
The Llama family
Meta's Llama series of open-weights models has shaped the open-source LLM ecosystem:
- LLaMA 1 (Feb 2023) — research-only release that leaked, sparking the open-LLM movement
- Llama 2 (Jul 2023) — first commercial open-weights release
- Llama 3 (Apr 2024) — 8B and 70B parameter sizes
- Llama 3.1 405B (Jul 2024) — first frontier-class open model
- Llama 4 Maverick (2025) — current flagship MoE architecture
Strategy
Meta has positioned open-weights as a competitive moat against closed-API rivals like OpenAI and Anthropic. By releasing weights freely, Meta:
- Commoditizes the underlying model layer
- Builds developer mindshare
- Avoids paying inference fees to a third-party
- Powers its own products (Meta AI in Instagram/WhatsApp/Facebook) at near-zero marginal cost
Compute investment
Meta announced US$60–65 billion in 2025 capital expenditure, much of it for AI infrastructure including its own MTIA training chips.
Latest news from Meta
8 most recent · auto-synced from RSSMeta is back in the LLM game after a year-long break - understandingai.org
Meta is back in the LLM game after a year-long break understandingai.org
Meta launches its most powerful AI tool Llama 3, know about it - Sangri Today
Meta launches its most powerful AI tool Llama 3, know about it Sangri Today
Latest activity
Full history →- 0
Meta: Llama 3 70B Instruct released
Meta: Llama 3 70B Instruct discovered via OpenRouter. Provider: Meta.
Llama 3 70B Instruct - 0
Meta: Llama 3 8B Instruct released
Meta: Llama 3 8B Instruct discovered via OpenRouter. Provider: Meta.
Llama 3 8B Instruct - 0
Releases timeline
Showing 12 most recent- Llama 3.3 70B Instruct
The Meta Llama 3.3 multilingual large language model (LLM) is a pretrained and instruction tuned generative model in 70B (text in/text out). The Llama 3.3 instruction tuned text only model...
- Llama 3.2 11B
Multimodal 11B model from Meta supporting text and image inputs.
- Llama 3.2 1B Instruct
Llama 3.2 1B is a 1-billion-parameter language model focused on efficiently performing natural language tasks, such as summarization, dialogue, and multilingual text analysis. Its smaller size allows it to operate...
- Llama 3.2 3B
Small on-device model for edge and mobile deployments.
Active models
- Llama 3.3 70B InstructActive
The Meta Llama 3.3 multilingual large language model (LLM) is a pretrained and instruction tuned generative model in 70B (text in/text out). The Llama 3.3 instruction tuned text only model...
131K ctx - Llama 3.2 1B InstructActive
Llama 3.2 1B is a 1-billion-parameter language model focused on efficiently performing natural language tasks, such as summarization, dialogue, and multilingual text analysis. Its smaller size allows it to operate...
60K ctx - Llama 3.2 3B InstructActive
Llama 3.2 3B is a 3-billion-parameter multilingual large language model, optimized for advanced natural language processing tasks like dialogue generation, reasoning, and summarization. Designed with the latest transformer architecture, it...
Community ratings
Rate Meta
Sign in to rate and review.
Comments
Sign in to leave a comment.