[Grok 4.5 Tops Tencent Hunyuan Agent Endurance Rankings, Over 60% of Tasks Completed Without Model Completion] According to monitoring by Dongcha Beating, Tencent's Hunyuan team, in collaboration with multiple universities, has released LHTB, a benchmark specifically designed to test whether AI Agents can continuously complete complex tasks on terminals. The benchmark includes 46 tasks spanning fields such as software engineering, experiment reproduction, scientific computation, and multimodal analysis. Each task runs for a maximum of 90 minutes and typically requires hundreds of consecutive steps. LHTB does not solely evaluate final success or failure. Partial task completion is also scored based on actual progress. Hidden validators check files, test results, and program outputs—self-reported completion by the Agent does not count. The first version of the paper tested 15 models. GPT-5.5 completed 7 tasks, ranking first. On average, each model consumed approximately 9.9 million tokens per task and took about 85 minutes. The currently published results have expanded to 20 models. Grok 4.5 topped the rankings with an average score of 0.505, completing 13 tasks. Out of the 46 tasks, 29 remain without any model meeting the completion standard. Most Agents can complete partial steps but often either exhaust the 90-minute limit or stop prematurely before finishing the task, failing to fully accomplish complex tasks. [Original Link]