Reasoning models¶

Research cut-off date: 2026-07-22. Status: Phase 2, not yet written. This file states the scope of the work, the arguments it owns, and the research required to complete it. It contains no findings, because none have been sourced. See the changelog for phase status.

Scope¶

Examines models that expend additional inference compute before answering. Covers chain-of-thought prompting, trained reasoning behaviour, hidden versus visible reasoning traces, caller-selectable reasoning effort, and the accuracy-per-token and accuracy-per-second trade-offs these create. Answers research question RQ4: whether reasoning-mode inference improves accuracy enough to justify its token and latency cost, and under what task conditions.

Arguments this chapter owns¶

Other files link here rather than restating these. Duplicated argumentation is a defect under the writing standards, not redundancy for the reader's convenience.

The definition of reasoning tokens and the taxonomy of reasoning modes.
The accuracy-versus-cost trade-off analysis for extended reasoning.
The evidentiary problem created by hidden reasoning traces, which prevent a caller from auditing what they are billed for.

Developed elsewhere¶

Cost arithmetic: 15-token-economics.md
Latency decomposition: 14-latency-and-throughput.md
Reasoning benchmark construction: benchmarks/knowledge-and-reasoning.md

Research checklist¶

Cite the primary sources for chain-of-thought and for process supervision.
Record, per profiled family, whether reasoning tokens are visible, whether they are billed, and at which price, each with a source and date.
Locate at least one Grade A study measuring accuracy as a function of reasoning token budget on a fixed benchmark, or record that none was found.
Compute the reasoning overhead ratio for each family where token counts are available, using scripts/calculate_token_costs.py.
State explicitly which task classes show no measured benefit from extended reasoning, with sources.

Completion criteria¶

This chapter is complete when every checklist item above is closed, when every numerical claim carries a footnote resolving to data/sources.csv, when every claim about a current model carries an absolute date, and when the twelve-point quality-control checklist in the research methodology passes.