Llama 3.3 70B
70B instruct model that approaches Llama 3.1 405B quality on many benchmarks at a fraction of the inference cost.
Post-training gains let 70B beat or match 405B on IFEval, MATH and code, while trailing on MMLU; 128K context. The last big text Llama before Llama 4.
- Date
- Friday, 6 December 2024
- Lab
- Meta
- Kind
- open-weights
- Access
- open weights (restricted license)
Figures
| Measure | Value | Measured by |
|---|---|---|
| IFEval | 92.1 vs 88.6 for Llama 3.1 405B per model card; MMLU Pro 68.9 | company |
Sources
This record was checked against its sources on 6 October 2026. How we check