Village: open-weight models are the best buy at 3% of budgets, down from 41% in June

Village, which publishes the LLM Atlas, reports that closed models now hold almost every point on the price-performance frontier. By its estimates, an open-weight model is the strongest model at 3% of budgets, nearly all at the cheap end (Google's Gemma 4 E4B and Xiaomi's MiMo-V2.6-Flash). In June the share was 41%, when DeepSeek V4 Flash alone led nearly a third of budgets. Village says today's share is the lowest it has measured; at each quarter's end since 2024 it was at least 6%.

Village credits OpenAI's GPT-6 Luna (September) and GPT-6.1 Terra (October), along with Anthropic's Claude Haiku 5.5, for taking back the budgets open models used to win. It offers three possible explanations, which it does not test: closed labs can distill their own frontier models, release whole families of models, and may serve more cheaply at scale.

The study takes the most capable model at 201 budgets spread on a log scale from 1% to 3x the cost of GPT-5. Capability scores are Village's estimates from public benchmark results, with costs from what evaluators paid through hosted APIs, as of October 31. Village notes that open models can be cheaper on your own hardware, and that widening the budget range to 0.3%–10x raises the open share to 16%.