open-sourceMeta·Released 2024-12
Llama 3.3 70B
Meta's open-weight workhorse — the default choice when you need an open model you can host, fine-tune, or air-gap.
Pricing verified August 21, 2026 · source
Key specifications
| Vendor | Meta |
| Tier | open-source |
| Context window | 128K tokens |
| Pricing | Open weights (free to self-host); hosted pricing varies by provider |
| Released | 2024-12 |
| Parameters | 70B |
| Categories | Text GenerationCode GenerationAdvanced ReasoningOpen Source |
Note: llama.com now redirects (301) to developer.meta.com/ai/. The homepage lineup shows only Llama 4 and Llama 3 tiles, no dedicated Llama 3.3 card, but the model card at developer.meta.com/ai/docs/model-cards-and-prompt-formats/llama3_3/, sourced from the GitHub MODEL_CARD.md, confirms 70B parameters, 128k context length, and the Llama 3.3 Community License Agreement. Facts unchanged from prior verification.
Key features
- •Open weights — run on your own hardware
- •Claimed performance close to Llama 3.1 405B
- •128K context window
- •Available via Together, Groq, Bedrock, and local runtimes
Compare with similar models
Model lineups change often. AI Flash Report tracks new model releases and maintains a deprecation calendar for models being retired.
← Back to all AI models