AI Release 📌 Let's discuss: YandexGPT got stuck in a loop for 37 minutes and burned 300,000 tokens with no result We put YandexGPT on a real agentic task — and the model entered an infinite loop: 295 repeated tool calls, 296 context rebuilds, 37 minutes and ~300,000 tokens. Zero useful output; we stopped it manually. The main question: the service has no protection against its own looping, while billing charges for every step. Everything falls on your bill, not theirs. By the way, according to the aggregate tests: 225K input + 161K output and 5.6 million cache-read tokens — with zero tasks solved. The full breakdown with numbers and a comparison with industry giants is on our website: YandexGPT got stuck in a loop: breakdown Have you caught YandexGPT looping? Share your experience in the comments — we'll put together a collection. AI Release YandexGPT got stuck in a loop for 37 minutes and burned 300,000 tokens with no result We don't test LLMs on pretty pictures but in the field — we put the model on a real task and see what it delivers. Next up was YandexGPT. The test result turned out not to be about sk ❤ 1 👍 1 3 views 10:18
Source: t.me · post in Telegram