Llama 4 Maverick vs Llama-3.1 Nemotron Ultra 253B

Llama 4 Maverick

Meta AI

Llama 4 Maverick, released by Meta on April 5, 2025, is a natively multimodal Mixture-of-Experts large language model with 400 billion total parameters and 17 billion active parameters per inference. It features a 1M token context window and supports text and image input, enabling strong performance on both language and vision tasks. Maverick targets open-source deployments requiring large-scale multimodal reasoning, released under Meta's custom license.

Llama-3.1 Nemotron Ultra 253B

NVIDIA

Llama-3.1-Nemotron-Ultra-253B-v1 is a 253-billion-parameter model from NVIDIA, derived from Meta's Llama 3.1 405B using neural architecture search (NAS) compression combined with NVIDIA's Nemotron post-training pipeline, which recovers and exceeds the base model's capability after structural compression. Released in April 2025, it supports toggling between a standard instruction mode and an extended reasoning mode via system prompt, allowing the same model to handle both rapid responses and deliberate chain-of-thought tasks. It is the flagship of the Nemotron family, available open-weight on HuggingFace and through NVIDIA NIM for enterprise inference.