nvidia is building a 1-trillion-parameter open model to beat deepseek and qwen

@_pheebini, nvidia reporter at @theinformation, broke the story: nvidia's nemotron 4 has one directive – be the best open-source ai model in the world. @ctnzr, one of the vps driving the effort, called it essential to nvidia's future. the nemotron coalition – @MistralAI, @perplexity_ai, @reflection_ai – launched in march to accelerate the push with shared data and compute.

the tension is built in: @nvidia's biggest gpu buyers are openai, microsoft, spacex. building a frontier model means competing with your own customers.

but nvidia's bet is that open-source grows the pie. more builders on open models means more builders on nvidia gpus. and right now, the openrouter leaderboard tells the story – chinese open models dominate the top 5 by volume. no major american open model is close.

nemotron 4 is nvidia's answer – not a policy response or a coalition statement, but a trillion-parameter model targeted for late fall 2026.

full segment via link. source: @theinformation