Models

Meta Llama 3.1 8B Instruct — a widely deployed small open-weight model.

Meta's Llama 3.2 1B Instruct — smallest open-weight Llama 3.2 text model.

Meta's Llama 3.2 3B Instruct — small open-weight text model for lightweight and on-device use.

Widely deployed open-weight model with strong general capabilities.

Largest open Llama 4 with 128 experts. 400B total, 17B active. Beats GPT-4o on benchmarks.

Natively multimodal MoE model with 10M context. 109B total, 17B active. Fits single H100.

Meta's Muse Glimmer 30B — compact Muse model with image input.