Meta

Comparing 8 models from Meta.

Model Total ▼ Released Context SizeCreative writingRule followingUtilityMathematicsToolingLanguageLogic
Llama 3.1 405B75.12%Jul 23, 24128k405B41.45%72.69%78.00%100.00%76.15%91.54%77.50%
Llama 3 70B74.93%Apr 18, 248k70B35.08%68.58%82.67%100.00%80.00%78.95%83.13%
Llama 3.1 70B70.11%Jul 23, 24128k70B40.65%65.35%76.39%100.00%70.90%74.29%76.25%
Llama 3.2 90B (Vision)69.39%Sep 25, 24128k90B41.62%64.63%75.00%95.00%68.85%75.72%76.88%
Llama 3.2 11B (Vision)66.35%Sep 25, 24128k11B27.10%63.08%71.56%100.00%64.74%71.57%83.13%
Llama 3.1 8B60.22%Jul 23, 24128k8B27.28%61.81%64.67%100.00%50.38%59.13%80.00%
Llama 3.2 3B51.19%Sep 25, 24128k3B30.33%60.89%47.22%10.00%48.97%60.77%54.37%
Llama 3.2 1B29.50%Sep 25, 24128k1B16.51%43.42%22.00%5.00%26.15%41.69%23.75%
Model Performance
Cost vs Performance

Compares total benchmark cost against overall score for Meta models. Quadrant lines are drawn at the median values.