Aider, Claude Code, and OpenClaw ran an identical model. Token use varied 70-fold.
When teams price an AI coding agent, they tend to scrutinize the model. But three recent benchmarking efforts suggest the The post Aider, Claude Code, and OpenClaw ran an identical model. Token use varied 70-fold. appeared first on The New Stack .
The human-written code generation agents Aider, Claude Code, and OpenClaw ran an identical model across different harnesses. Token usage varied significantly, ranging from approximately 3,500 to 292,000 tokens per task. This disparity was primarily due to the harness software, which includes the system prompt, tool descriptions, and environment setup, requiring large blocks of text to be processed on every inference request.
The upfront cost of this "startup tax" could be significant, with token usage for the harness sometimes exceeding 26,000 tokens, even when only a small portion of the tokens are actually used for the task. Despite this, developers should prioritize optimizing the prompt floor and turn count before considering more complex optimizations.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
