I use 30 million tokens a day on average for work—what level of consumption is that?
2 weeks ago
I suddenly noticed that our company's ChatGPT provides this data. Most of the time I use GPT 6.1 Sol.
I'm not a programmer—is my usage a bit low?
2 weeks ago
I suddenly noticed that our company's ChatGPT provides this data. Most of the time I use GPT 6.1 Sol.
I'm not a programmer—is my usage a bit low?
Is this data available only as a total, or can it be broken down by day or by model?
For non-development roles, this amount of usage isn't unreasonable, but the risk is that it's easy to treat model output as correct by default. The boundary is clear: looking things up, drafting, and organizing are fine, but anything involving numbers, conclusions, or accountability still needs to be reviewed yourself.
The differences between roles are actually quite significant. People writing code or doing data analysis paste in long contexts at once, so tokens add up quickly; those in admin or marketing mostly have short conversations, which look small in volume but happen frequently. On top of that, a company's billing basis and whether system prompts are included also affect this number, so comparisons between teams in different regions may not be apples-to-apples.
The hard part is converting token volume into actual output.
Many people only look at usage and ignore caching and repeated questions—they keep opening new conversations for the same question. Usage goes up, but efficiency may not. Also, if your company caps usage by seat, getting blocked at the end of the month is more of a hassle than using a little extra at other times. It's best to figure out the quota rules first.
I'd like to confirm whether this 30M only counts your personal account, or if it also includes calls from the shared workspace. Also, is GPT 6.1 Sol the company-wide default model, and is it possible that calls from other tools in the backend are being counted too? Without these details clear, any judgment is a bit shaky.
Thirty million tokens a day sounds like a lot, but what really matters is the time cost. If that volume is concentrated in a few long conversations with long documents pasted in over and over, the time spent waiting for responses and manually checking them could outweigh the savings. Conversely, if it's a large number of short questions and answers, the efficiency gains will be more obvious. So looking at the total alone makes it hard to judge how much is enough—it depends on how it's used.