Open-Source AI Struggles to Close Gap with Closed Rivals

Over the past week, artificial intelligence developers have posted rave reviews on X and Reddit about the quality of Nemotron, a series of large language models Nvidia built using Meta Platforms’ open-source Llama model.

The buzz about Nemotron rekindled questions about whether the race between open- and closed-source models is tightening. The short answer seems to be no—but not necessarily because there’s a big difference in quality between closed and open models.

To be sure, closed models from OpenAI or Anthropic are still rated somewhat better than their open-source counterparts, according to the Lmsys leaderboard and plenty of anecdotal evidence. But what is also keeping these closed models ahead is their ease of use.