GPT-5.6 Batch Processes 78K Classifications Overnight for $60

GregKamradt · x · 2026-08-01

Developer Greg Kamradt shared a case study of massive AI task processing. He used GPT-5.6 Luna's batch mode to run overnight, completing 78,000 different classifications.

The task consumed 143 million input tokens across 158,000 requests. Leveraging the low batch mode rate of $0.10 per million tokens, the total cost was only $60. This demonstrates the extreme cost-effectiveness of batch processing APIs for large-scale structured tasks.

Original post →

More from coding & agent

coding & agent channel →