Nvidia introduces open-source NVLM 1.0 models

Nvidia has officially entered the ring with a powerful open-source AI model, NVLM 1.0, challenging industry giants like OpenAI and Google.

The company’s new NVLM 1.0 family of large multimodal language models promises to deliver cutting-edge capabilities across both visual and text-based tasks.

Leading the pack is the 72 billion parameter NVLM-D-72B, a model designed to perform at the highest level, making a massive impact on vision-language tasks while improving traditional text-based outputs.

The release of NVLM 1.0 marks a notable shift in the AI ecosystem, which proprietary models have largely dominated. Nvidia’s decision to make these model weights publicly available—and eventually release the training code—offers researchers and developers access to tools that rival the likes of GPT-4. This is a rare move in an industry where most advanced models remain under lock and key, tightly controlled by tech giants.

As Nvidia stated in their research paper, “NVLM 1.0 achieves state-of-the-art results on vision-language tasks, rivaling both proprietary and open-access models.”

What this means for developers is a new frontier in AI accessibility, much like what Meta did with Llama 3.2, giving smaller labs and independent researchers a chance to work with top-tier AI tools without having to navigate the often prohibitive costs or corporate restrictions.

The open-source release of NVLM 1.0 has generated excitement across the AI research community. One prominent researcher highlighted the significance of the model on social media, stating:

Wow nvidia just published a 72B model with is ~on par with llama 3.1 405B in math and coding evals and also has vision