Nvidia is reportedly training a trillion-parameter open-source model, Nemotron 4, possibly ready by late fall

消息称英伟达开发万亿参数开源 AI 模型 Nemotron 4，目标挑战全球顶级

The Information reports, citing project participants, that Nvidia is building the Nemotron 4 series, with the largest model reaching at least 1 trillion parameters. Nvidia VP of GenAI Kari Briski said in an email that the investment is driven by the belief that 'every company and every country needs accessible frontier open-source models.' Final training isn't done yet, but employees think it could be ready by late fall. Nvidia is one of the few US big-tech firms actively releasing open-weight models, aiming to broaden AI adoption and boost demand for its GPUs. The post does not disclose benchmark scores or pricing.

Why it matters: Nvidia building a trillion-parameter open-source model with concrete specs and an internal timeline — not a vague rumor. The VP's email quote about 'every company, every country needs frontier open-source models' gives this a clear strategic framing. Score held below 85 becaus...