Fable 5.1 burns 102K reasoning tokens and hits the 128K output ceiling mid-code
rohanpaul_ai · x · 2026-09-03
- Setup: @thehypedotnews gave Fable 5, Fable 5.1, and Opus 5 one reference image each and asked them to procedurally rebuild Lord of the Rings landmarks (Bag End, Barad-dûr) in Three.js — single HTML file, no meshes/textures/assets, just Three.js from a CDN, run through OpenRouter, with three camera shots per scene on keys 1/2/3.
- Results: Fable 5 was cheapest and fastest; Fable 5.1 stood out on visual detail.
- Key finding: Fable 5.1 spent 102,116 tokens reasoning and hit its output ceiling before finishing the file. Anthropic documents a 128K max output for Fable 5.1, and current-turn thinking counts against the same limit — heavy reasoning can leave too little room for code, truncating large codebases.
- Takeaway: watch the thinking/output budget trade-off when generating big files in one turn.
More from coding & agent
- mcp-doctor: a static analysis linter for MCP servers that surfaced real bugs in 12 popular repos — Wide_Imagination_970 · 2026-09-03
- Stripe launches Directory preview to help AI agents find and use external providers — jeff_weinstein · 2026-09-03
- Akkru launches financial data MCP so AI agents can trace numbers back to filings — camiggggg · 2026-09-03
- Matt Shumer explains Fable's architecture: locked world structure lets versions share one world — mattshumer_ · 2026-09-03
- GLM-5.3-Flash priced at 0.06x: Factory confirms all-day droid usage without rate limits — matanSF · 2026-09-03
- Qwen 3.8 users report better results forcing reasoning effort to low, llama.cpp flag shared — Jorlen · 2026-09-03