Addy Osmani's Claude model picking guide: Opus 5.5 default, Fable 5.2 costs 2.8x more
udmrzn · x · 2026-11-16
Addy Osmani published a guide to choosing among Claude Haiku 5.5, Sonnet 5.5, Opus 5.5 and Fable 5.2. Core advice: start with Opus 5.5 at medium effort — it costs half of Fable 5.2 per token; raise effort before switching models.
Tiering: drop to Sonnet 5.5 for well-scoped everyday work; use Haiku 5.5 for classification/routing/extraction and subagents; reserve Fable 5.2 for major migrations and long unattended runs. Citing garry tan's benchmark: across 52 coding tasks in Claude Code 2.1.322, Fable 5.2 solved about as many as Opus 5.5 (82.7% vs 84.6%) but cost 2.80x more per solved task.
More from coding & agent
- Claude Code's full system prompt leaked: 7,832 lines, 128 tools, 44 skills on GitHub — udmrzn · 2026-11-16
- Forensics on FBI-flagged ARTEX pentest framework: deleted auth guard, backdoored Docker image — udmrzn · 2026-11-16
- Dev floats an MCP mapped to physical addresses so agents could query residents and businesses — mobileraj · 2026-11-16
- GitHub's new Copilot app brings repos, PRs and issues together — 0xkarasy · 2026-11-16
- Building AI agents is the easy part — developers debate what blocks revenue — SufficientLab218 · 2026-11-16
- Three practical tips for getting more out of Fable 5.2, from permission-giving to orchestrating Opus 5.5 agents — dreamwieber · 2026-11-16