# Fable 5.1 对比 Fable 5 与 Opus 5：单图重建《魔戒》3D 场景编码实测

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-09-03 06:43
- AIHOT 分数：60
- AIHOT 链接：https://aihot.virxact.com/items/cmtkplqlf01hbroals8nfxiqg
- 原文链接：https://x.com/rohanpaul_ai/status/2095281428741415167

## AI 摘要

@thehypedotnews 用 Three.js 让 Fable 5.1、Fable 5 和 Opus 5 分别从单张参考图程序化重建 bag end、barad-dûr、rivendell 三个《魔戒》场景，全程经 OpenRouter 运行。

## 正文

Fable 5.1 vs fable 5 vs opus 5

cool experiments by @thehypedotnews

the test gave Fable 5, 5.1, and Opus 5 the same reference scenes and asked each to recreate them procedurally from scratch in Three.js.

Fable 5 was both the cheapest and the fastest, but fable 5.1 stood out on visual detail.

This test also exposed a token constraint because Fable 5.1 spent 102,116 tokens reasoning and hit its output ceiling before finishing the file.

Anthropic documents a 128K maximum output for Fable 5.1, and current-turn thinking counts against that same limit, so heavy reasoning can leave less room for code. If the model spends too much of it thinking, it can hit the 128k limit before finishing a large codebase.

### 引用推文

> thehype.：fable 5.1 vs fable 5 vs opus 5 – three lord of the rings landmarks, built in 3d from one image the setup: one reference image per scene, one html file per build...