Anthropic says start most coding work on Opus 5.5, not Fable 5.2 — effort beats model size
blaizedsouza · x · 2026-11-15
Anthropic's Fable 5.2 launch post advises that most coding and knowledge work should start on Opus 5.5, not the bigger Fable. A larger model doesn't mean better results — the usual problem is missing context or a wrong effort setting; a bigger model on a bad prompt just costs more for the same answer. Only switch up when the model had everything it needed and still failed.
The attached 25-task Claude Model Selection Guide maps tasks to models and effort:
- Everyday: quick answers/summarizing → Haiku 5.5 ($0.10/$0.50); drafting → Sonnet; reading screenshots → Opus
- Code: small bugs → Sonnet HIGH; large-codebase refactors → Opus HIGH; 30+ min coding agents → Opus EXTRA HIGH
- Agents: bulk processing and subagent grunt work → Haiku LOW; hours-long autonomous runs → Fable HIGH
- Weak answer? Add context → raise effort one step → only then move up a model
The poster also cut a team to a $30 bill by fixing one mode with 'Viktor' and picking models only when a job demands it.
More from coding & agent
- Cutting a Research Agent From 12 Tools to 3 Boosted Success Rate From 52% to 84% — Such_Reflection8996 · 2026-11-15
- Bypassing Opus 5.5 refusals with a 'boss' agent: multi-agent setups revive the confused deputy flaw — rickasaurus · 2026-11-15
- NVIDIA open-sources SkillEvaluator, a multi-tier framework for evaluating AI agent skills — NVIDIA · 2026-11-15
- Claude-BugHunter: Claude Code skill bundle with 82 skills for bug hunting hits 5.9k stars — elementalsouls · 2026-11-15
- Three shifts since spring: one heavy user's 'How I use AI in November 2026' — TheTuringPost · 2026-11-15
- Every team builds a video editor and writing coach with Claude Fable 5.2 — every · 2026-11-15