Llama 3.3 70B
Language modelPrevious generationOpen weightsCard summarizedThe Llama 3.3 instruction tuned text only model is optimized for multilingual dialogue use cases
Open weights (static model, still published)● Re-checked 12 Oct 2026 · 10 facts · 0 correctionsSource: Llama 3.3 Model Card (MODEL_CARD.md)
Use it via API
Where to call it
- Self-host (weights)
meta-llama/Llama-3.3-70B-Instruct - Amazon Bedrock
Llama 3.3 70B - Together AI
meta-llama/Llama-3.3-70B-Instruct-Turbo
Good for
- Commercial and research use in multiple languages
- Assistant-like chat (instruction-tuned)
- Using model outputs to improve other models, including synthetic data generation and distillation
Overview
Llama 3.3 70B
The Llama 3.3 instruction tuned text only model is optimized for multilingual dialogue use cases
- Cost
- $0.10 host
- Context
- 128K
- Cheapest host
- $0.10 / $0.32
- Max output
- —
- Knows up to
- December 2023
- Released
- 6 Dec 2024
Llama 3.3 is a 70B instruction-tuned, text-only model from Meta, optimized for multilingual dialogue. It uses an optimized transformer with grouped-query attention, a 128K context and 15T+ pretraining tokens; on Meta's own table it matches or comes close to Llama 3.1 405B on several benchmarks. Model Information; Benchmarks
meta-llama/Llama-3.3-70B-Instruct- Input / output per 1M
- $0.10 / $0.32 via DeepInfra
- Context window
- 128K
- Max output
- —
- Knows up to
- December 2023
- Reasoning
- —
- Weights
- Open · Llama 3.3 Community License
About this card. A model card is the maker's own account of a model: what it's for, how it was trained and tested, and where it falls short. This page condenses Meta's Llama 3.3 Model Card (MODEL_CARD.md). Every claim carries its source page.