Even more striking: ~50% of requests already exceed 128k tokens. The driver isn't user prompts getting longer. It's ever…

SemiAnalysis Twitter · SemiAnalysis (@SemiAnalysis_) · 2026-05-22

SemiAnalysis reports that approximately 50% of real-world AI inference requests already exceed 128k tokens, with the growth driven by agentic prefill — system prompts, tool definitions, MCP schemas, and prior context — rather than longer user messages.

Open original ↗

Appears in

Agentic Workloads Rewriting LLM Inference Economics

Extraction

Topics: agentic-workloadscontext-lengthinference-economicstoken-usage

Claims

Roughly 50% of AI inference requests already exceed 128k tokens in real-world usage.
Token growth is driven by agentic prefill context — system prompts, tool definitions, skills, MCP schemas, and prior turns — not by users typing longer messages.
Agentic workloads fundamentally change the economics of AI inference at scale.

Key quotes

The driver isn't user prompts getting longer. It's everything the agent stuffs in before you even type: system prompts, tool definitions, skills, MCP schemas, prior turn context, file contents.