I just recently started using gemini-3-flash-preview in a project of mine and have been happy with the results and costs. However, beginning February 4th, the costs for “Generate_content text output token count for gemini 3 flash” have exploded by about 40x!
My daily usage is very consistent and the tool I use it for always performing the same task, so I’m not sure what’s happening here. There have also been no changes to the tool that would explain this behavior. Kindly help me out, as this makes Gemini unattractive from a performance/cost perspective.
yeah, somethings up last time there was excessive cost to actual token usage was when gemini 3 launched. testing same app at 1/20th of the usage over the past couple of days and the cost discrepancy vs the cloud api logs is HUGE! the spike at 12/17 is gem 3 flash release.
The same thing happened to our project.
I even suspected that the issue was caused by the Key being leaked or something of the sort, because the costs didn’t match the usage and they were consistent.
So to be sure, I created a new project, only for recreating the API key, restricted it to a single IP, the usage in the dashboard matches our usage but the costs got even higher.
