Traictory tracks 14 models from Meta. The earliest entry is Llama 3.1 405B Instruct from July 2024; the most recent is Muse Spark 1.3, released in September 2026. 5 of them shipped in the last twelve months.
The strongest Meta model in our data is Llama 3.3 70B Instruct, with an average benchmark score of 79.9% across 9 published results, including 50.5% on GPQA and 86.0% on MMLU. Across the 14 Meta models with benchmark data the average is 69.5%.
12 of the models are available through an API, with input pricing from $0.01 to $3.50 per million tokens; Llama 3.2 3B Instruct is the cheapest. The largest context window is 10M tokens on Llama 4 Scout. 9 of the 14 models accept images or other non-text input. 10 models ship with downloadable weights; the rest are API-only.
Meta Models
Meta Model Comparisons
Other Companies
.svg?dpl=dpl_3EZdmE8NdzRwHesETrvSRCczw81T){kind=icon}
Meta model catalog updated in 2026. See model comparisons and best models for coding.