Metrics

Codex reports token usage as cumulative totals in the usage object carried by turn.completed. The harness layer reads those totals from the JSON stream and produces the normalized token classes. Aggregation is Last: each usage-bearing line replaces the running total, so the final reported totals are the ones recorded.

Token classes

Normalized class	Codex key
Uncached input	input_tokens less cached_input_tokens
Cached input	cached_input_tokens
Output	output_tokens
Reasoning	reasoning_output_tokens

Codex’s input_tokens is inclusive of cached reads, so the cached input is subtracted from it to yield the uncached input. Codex has no cache-creation class. Reasoning is reported on its own key and tracked separately from output_tokens, which already excludes it.

Cost

Codex reports no cost of its own, so the comparable cost is computed from the model’s list price applied to the recorded token classes. The bare OpenAI model ID resolves to the catalog entry that lists it as a Codex alias, so gpt-5.5 is priced from the entry carrying the gpt-5.5 alias. The billed-rate lookup prefixes it with openai/, so gpt-5.5 is observed on OpenRouter as openai/gpt-5.5. See Metrics for that contract.