Agentic workloads are quietly rewriting inference economics. We pulled data from 432k real coding agent requests at Semi…
SemiAnalysis Twitter · SemiAnalysis (@SemiAnalysis_) · 2026-05-22
SemiAnalysis analysis of 432,000 real coding agent requests finds a median input size of 96k tokens — larger than the full text of The Great Gatsby — revealing that agentic AI workloads are dramatically exceeding commonly assumed context sizes.
Appears in
Extraction
Topics: agentic-workloadscoding-agentsinference-economicstoken-usage
Claims
- The median coding agent request in SemiAnalysis's dataset of 432k requests is 96k input tokens.
- This median far exceeds commonly assumed values of 32k or 64k tokens.
- 96k tokens is comparable in length to the entire text of The Great Gatsby.
- Agentic workloads are quietly but significantly reshaping AI inference economics.
Key quotes
The median one isn't 32k, isn't 64k, but 96k input tokens. For context, that's more than the entire text of The Great Gatsby being shoved into the [model].