The A.I. Daily
Report: closed models dominate price-performance frontier as open share hits low
Village's LLM Atlas shows open-weight models are the best buy at only about 3% of budgets, the lowest measured, versus at least 6% at each quarter-end since 2024. The share hit 41% in June on DeepSeek V4 Flash, before OpenAI's GPT-6 Luna, GPT-6.1 Terra and Anthropic's Claude Haiku 5.5 undercut those budgets.