Llama 3.3 70B
Self-hosted on-premise deployments and edge dense inference.
SWE-bench Verified
38.8%
MMLU-Pro
68.3%
MATH-500
75.8%
Context Window
128,000 tokens
| Specification | Value |
|---|---|
| Developer / Lab | Meta AI |
| Parameters Total | 70.6B |
| Parameters Active | 70.6B (Dense) |
| Architecture Type | Grouped-Query Attention (GQA) Dense Transformer |
| License | Llama 3.3 Community License |
| Input Pricing (per 1M) | $0.15 |
| Output Pricing (per 1M) | $0.60 |