NVIDIA Corporation (NVDA.US) unveiled its self-developed large model Nemotron 3 Ultra: cost-effective, with reasoning costs reduced by 90% compared to closed-source models.

Notice that NVIDIA Corporation (NVDA.US) stated on Wednesday that its open-source Nemotron 3 Ultra AI model has surpassed closed-source models in terms of cost. In a statement, NVIDIA Corporation disclosed that the open-source framework Langchain reported its Deep Agents suite, specifically designed for Nemotron 3 Ultra, is able to achieve inference costs 10 times lower per run compared to some "leading" closed-source models, completing more tasks and achieving higher throughput. NVIDIA Corporation added that when compared to LangChain's Deep Agents benchmark tests, Nemotron 3 Ultra also performed on par with the highest-scoring models in business tasks. Co-founder and CEO of LangChain, Harrison Chai Si, stated during a press conference, "The method of building better intelligent agents is to continuously improve the systems around the model. When teams can put them together for fine-tuning, memory, tool usage, evaluation, and model behavior will have a compound effect. Our collaboration with NVIDIA Corporation shows that businesses can obtain strong performance from open-source technology stacks while maintaining control over the intelligent agent systems they are building." NVIDIA Corporation also noted that Abridge, Amdocs, and Box are embedding specialized intelligent agents directly into their platforms, and Ernst & Young (EY) is expanding the application of NVIDIA Corporation's capabilities around its NemoClaw blueprint. The "ultimate moat" of software and hardware integration NVIDIA Corporation's strategy of developing large models is not aimed at directly competing with OpenAI or Anthropic by selling large model APIs, but rather to strengthen and expand its dominant position in the AI industry chain. NVIDIA Corporation's biggest moat is not the chip itself, but the CUDA software ecosystem. By introducing its own large models like Nemotron 3 Ultra and packaging them into the NVIDIA NIM (microservice framework), enterprises can deploy and fine-tune them with a single click on NVIDIA Corporation hardware. NVIDIA Corporation's development of large models serves to make its software stack (AI Enterprise) more attractive, thereby firmly tying customers to NVIDIA Corporation's hardware and software ecosystem, known as "full-stack monetization." Closed-source large models like OpenAI must be accessed through public clouds, which many government agencies, defense contractors, financial giants, and sensitive industries cannot use due to data privacy and security restrictions. NVIDIA Corporation, by launching the "weight open-source, allow private deployment" Nemotron model, collaborates deeply with enterprise software giants like Palantir and DataRobot. These large models, which run in a completely isolated and secure local environment, are able to effectively penetrate the high-profit government and enterprise market that public cloud giants cannot reach. Just as Microsoft Corporation developed the Surface computer to showcase the Windows lineup, NVIDIA Corporation trains its own large models to demonstrate to global enterprises how the performance and cost of large models can be optimized to the extreme on Hopper and Blackwell chips. For instance, with the latest Nemotron 3 Ultra, NVIDIA Corporation, in collaboration with LangChain, has proven that its inference costs are nearly 10 times cheaper than popular closed-source large models in the market. Beyond general language models, NVIDIA Corporation has also deepened its presence in models like Cosmos (world model) and GR00T (Siasun Robot & Automation large model). These models can simulate the real physical world and mechanical reactions, serving as indispensable brains for autonomous driving, humanoid Siasun Robot & Automation, and intelligent factory digital twins (Omniverse). Traditional internet large model companies are not well-equipped for this industrial-grade hardcore AI race, which is where NVIDIA Corporation's chips will play a major role in the future.