Closed-source LLMs regain dominance on the cost-performance frontier

Data from Village's LLM Atlas indicates that closed-source models have regained dominance on the cost-performance frontier across all budget tiers, with open-weight models remaining the most cost-effective option at only 3% of budget points. This share plummeted from 41% in June, primarily because OpenAI's GPT-6 Luna and GPT-6.1 Terra, along with Anthropic's Claude Haiku 5.5, have captured price segments previously held by models such as DeepSeek.