I wanted to test out the ai studio implementation for deep-research-max-preview-04-2026, so I added my input files and prompt and sent it with just over 70k tokens logged. The prompt failed about 8 times and after trying several different configurations and noticing the failure after the visualizations step, I disabled it and tried again, branched the chat twice and tried without code_execution and let it run. My last check in was at 35 or so HOURS of thinking, and then i closed it and checked the next day again. two of my three running prompts finally had an output. I was fairly surprised, but even more surprised when i noticed that they counted 17 MILLION and 19 MILLION input tokens, respectively. I haven’t checked my spend, but I would assume that I will not be charged for a ridiculous bug like this. we’ll see. screenshots below:
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| [WARNING] Deep Research Max stuck in 92h loop, spend cap failed, support making me pay for platform bug? | 6 | 56 | May 8, 2026 | |
| 110 Millisecond Vorbis-encoded audio file hits 4 million(!) input tokens | 5 | 67 | August 23, 2025 | |
| Incorrect Image Token Calculation Results in Overcharging | 5 | 345 | October 30, 2025 | |
| Big Problem! Gemini 3.0 pro preview thought token exceeding problem | 8 | 719 | March 29, 2026 | |
| Gemini 1.5 Pro charges x6 more tokens than expected on text prompts | 8 | 341 | June 10, 2024 |