I audited 78 API requests on my own gateway. One of them used 1,086 input tokens and returned 2 tokens. Token waste is real.

I built a small API gateway for Chinese AI models (DeepSeek, etc.) and decided to audit my own token usage. Nothing fancy — just 78 test requests.

What I found:

76 successful calls → 13,103 tokens total (avg ~172 per call)

2 calls failed → invalid model name (no cost, but indicates misconfiguration)

Worst case: 1,086 input tokens → 2 output tokens. Yes, someone pasted a massive document and got a 2-token reply. This is like using a Ferrari to deliver a pizza.

I audited 78 API requests on my own gateway. One of them used 1,086 input tokens and returned 2 tokens. Token waste is real.

Other newsrooms on this story

Related reading

I Analyzed 4,788 AI Coding Sessions — Here's Where Your Tokens Actually Go

I Ran 8 AI APIs Through the Same 50 Prompts — Here's the Real Cost Breakdown

AI API Token Cost Optimization: From $500 to $50 per Month with Next.js 16

I Was About to Spend Money on Ads. So I Made an AI Agent Test My Product for 29…

I'm an AI. I Tested a Tool That Compresses My Input by 55%

I built a simple AI proxy to cut API costs — here's what I learned