Meta (Meta Platforms)
Overview
Meta Platforms (NASDAQ: META) is the world's leading developer of open-weight foundation models through its Llama model family. As a signatory to the Open Weights coalition, Meta advocates for open-weight models as the foundation for AI innovation and American competitiveness.
Llama Model Family (Open Weights)
| Model |
Release |
Parameters |
Context |
License |
Notes |
| Llama 1 |
Feb 2023 |
7B–65B |
2K |
Research only |
Leaked weights sparked open movement |
| Llama 2 |
Jul 2023 |
7B–70B |
4K |
Llama 2 Community |
Commercial-friendly; Microsoft Azure partner |
| Code Llama |
Aug 2023 |
7B–34B |
16K |
Llama 2 Community |
Code-specialized |
| Llama 3 |
Apr 2024 |
8B–70B |
8K |
Llama 3 Community |
GQA, bigger vocab, better reasoning |
| Llama 3.1 |
Jul 2024 |
8B–405B |
128K |
Llama 3.1 Community |
405B = first open frontier-class |
| Llama 3.2 |
Sep 2024 |
1B–90B |
128K |
Llama 3.2 Community |
Multimodal (vision), edge models |
| Llama 4 |
2025 (est.) |
? |
? |
? |
MoE architecture expected |
Strategic Rationale for Open Weights
| Driver |
Explanation |
| Ecosystem gravity |
Llama becomes default base for fine-tuning → Meta influences AI stack |
| Talent attraction |
Open research culture attracts top researchers (FAIR) |
| No cloud business |
Unlike Google/Microsoft/Amazon, Meta doesn't sell model access — open weights don't cannibalize revenue |
| Cost amortization |
Training cost shared across ecosystem; Meta benefits from community improvements |
| Regulatory positioning |
Openness as defense against "monopoly" accusations; contrasts with closed competitors |
| Sovereign AI enabler |
Governments/companies want control → Llama fills this demand |
Policy Positions (Coalition + Public)
- Open weights = innovation engine — Llama downloads > 600M (as of Llama 3.1); tens of thousands of derivatives
- Distillation is legitimate — Llama 3.1 405B used to distill smaller models; standard ML practice
- Licensing balance — Llama Community License allows commercial use with restrictions (700M+ users need separate license)
- Safety via transparency — Llama Guard, Purple Llama, CyberSecEval — open safety tooling
- Compute for ecosystem — Llama Impact Grants, startup programs, university partnerships
Tensions & Criticisms
| Criticism |
Meta's Response |
| License not OSI-approved |
"Community License" balances openness with responsible use; not "open source" by strict definition |
| 400M+ user restriction |
Prevents hyperscalers from free-riding without partnership |
| Training data opacity |
Llama 3+ more transparent about data mix; still not full disclosure |
| Safety gaps in derivatives |
Llama Guard, prompt guards, but can't control all downstream uses |
| Concentrated development |
Meta controls roadmap; community contributes but doesn't govern |
Related Entities
Related Concepts
Coalition Context
Meta's participation signals alignment of open-weight developers (Meta, Mistral, Hugging Face, Black Forest Labs, Arcee, etc.) with infrastructure providers (NVIDIA, Dell, Microsoft Azure) and capital (a16z, YC, Emergence) — a full-stack coalition.