DeepSeek's New Model Tops Agent Arena Open-Source Rankings

AI arena platform Arena announced that DeepSeek's newly released DeepSeek-V4.1-Flash-1017 has topped the open-source rankings on both Agent Arena and Code Arena: WebDev, ranking 11th overall. The results come from 7,498 real usage sessions, with an overall net improvement score of 5.13%. The model leads all open-source models in task success confirmation, user feedback, Bash command reliability, and steerability. Its task success score is 11.54%, ranked 8th overall. Its tool hallucination rate is just 0.48%, ranked 7th — a lower number means fewer made-up tools. On cost, the median cost per task is about 0.18 to 0.19 US dollars, or roughly 1.2 yuan. Arena also placed the model on the Pareto frontier of performance and cost, meaning no other listed model beats it on both at the same time.