Meta
Meta's 70B Llama 3 base model — a strong pretraining foundation for fine-tuning and research at large scale.
Meta's 70B Llama 3 instruction model — the most capable of the original Llama 3 release, strong on reasoning and chat.
gradientai's 8B Llama-3 instruct variant — optimized for low-latency inference on Gradient with strong conversational performance.
Meta's Llama 3 8B base model — efficient, Apache-licensed, strong baseline for fine-tuning.
Meta's 8B instruction-tuned Llama 3 model — fast, capable, and one of the most popular open models for chat and agents.
GAIA-LLM-Agent (Llama-3-70B, Northeastern University)
