---
schema: 1
type: term
title: _thinking traces_
description: >-
  **_Thinking traces_** are the visible reasoning a model produces on its way to
  an answer — the step-by-step output you can inspect when evaluating how a
  model reasons, not just what it concludes.
canonical: >-
  https://liquidgold.cc/channel/everyinc/explore/term/thinking-traces-c281723cae59
markdown: >-
  https://liquidgold.cc/channel/everyinc/explore/term/thinking-traces-c281723cae59.md
llms: null
parent: 'https://liquidgold.cc/channel/everyinc/llms.txt'
channel: everyinc
updated: '2026-10-23T22:12:35.346Z'
language: null
children: []
---


# \_thinking traces\_

Canonical: https://liquidgold.cc/channel/everyinc/explore/term/thinking-traces-c281723cae59

Parent: [Channel index](https://liquidgold.cc/channel/everyinc/llms.txt)

## Definition

**_Thinking traces_** are the visible reasoning a model produces on its way to an answer — the step-by-step output you can inspect when evaluating how a model reasons, not just what it concludes.

In the channel's material on hands-on model evaluation, **_thinking traces_** comes up in the context of testing how models reason. When the Every team runs day-zero checks on a new release — as with **Opus 4.8** — part of the exploration is understanding not just what a model answers, but the reasoning it shows on the way to an answer. That visible reasoning is what lets a reviewer judge where a model excels, where it hesitates, and how much reasoning effort a given task actually needs.