Open source AI is just 4 months behind closed frontier models — and 10x cheaper

Source and reference

source community rarely shies away from battle ( remember Windows and, now, Linux and Android?) with open advocates saying that proprietary models are merely an enterprise wrapper around the model that has memory, observability controls, routing intelligence, and connectors). That works out to be expensive wrapping paper if open-source models are roughly 10x less per token. Fearmongering factor from the big firms Boris Renski , founder & CEO at AI agent integration startup Apelogic , tells The New Stack that frontier labs are “hard at work pushing fear” about uniquely smart, dangerous models and imminent AGI. “That fear factor is designed to distract from the benchmarks that show open source models are only four months behind at a fraction of the cost,” Renski says. “That means most of the time companies are not paying OpenAI or Anthropic for intelligence, but for ‘commodity...