Nvidia has quietly introduced a new artificial intelligence model named Llama-3.1-Nemotron-70B-Instruct, which has outperformed OpenAI's GPT-4 and Anthropic's Claude3.5, marking a significant shift in the competitive landscape of the AI field. This model was released on the renowned AI platform Hugging Face, and despite minimal publicity, its exceptional performance quickly garnered industry attention.
According to Nvidia's reports, this new model achieved the highest scores in multiple benchmark tests, including 85.0 in the Arena Hard benchmark, 57.6 in AlpacaEval2LC, and 8.98 in GPT-4-Turbo MT-Bench. These scores have rapidly propelled Nvidia to become a new leader in AI language understanding and generation.
Nvidia is well-known for its GPUs, especially in driving the development of AI systems. Now, not only does it lead in hardware, but it also demonstrates formidable capabilities in AI software development. The release of this model also indicates that Nvidia is strategically expanding, challenging traditional language model development paradigms.
Nvidia improved Meta's open-source Llama3.1 model using reinforcement learning from human feedback (RLHF) technology. This method allows the model to better understand human preferences and provide more natural responses. This model not only performs excellently but also handles complex queries without additional prompts or specific markers.
To support businesses in exploring AI solutions, Nvidia offers a free hosted inference service compatible with OpenAI's API interface, making advanced AI technology more accessible and helping companies across various industries experiment and implement.
Despite this, Nvidia reminds users that there may be risks in applying this model to special fields such as mathematics or law, and companies should use it with caution to prevent errors or misuse.
With Nvidia's introduction of the new model, competition in the AI field is intensifying. This release marks a new phase in the competition for AI system development, and Nvidia's move will force other tech companies to reconsider their strategies and accelerate their R&D processes.
Project entry: https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Instruct-HF
Key points:
🌟 The newly released Llama-3.1-Nemotron-70B-Instruct model outperforms OpenAI and Anthropic, marking a significant change in AI competition.
🚀 Nvidia extends its AI development from hardware to software, showcasing its strong model development capabilities.
💻 The model provides an easy-to-use API interface, enabling more businesses to access and utilize advanced AI technology.