Prompting Claude Fable 5.2
What happened
The article outlines prompting patterns specific to Claude Fable 5.2, highlighting its new effort‑level system that controls how much the model thinks, makes tool calls, and uses tokens. Different effort settings—low, medium, high, xhigh, and max—are suited to various tasks such as quick back‑and‑forth queries, interactive coding, hard research, and open‑ended scientific problems. Users should recalibrate effort levels when migrating from earlier Claude models, as performance and token usage differ. The guide also advises handling long‑turn agentic work, subagent expectations, code ownership statements, and making final reports easy to follow. It stresses testing effort levels against your own evaluations to find the optimal balance.
Why it matters
For ordinary readers, understanding these effort controls helps developers and businesses get more predictable costs and response times when using Claude Fable 5.2. It also makes it easier to tailor the model’s behavior to specific workloads without unnecessary delays.
Reality check
An automated, sceptical read of the sourceIncremental. The changes refine existing prompting techniques and add configurable effort levels, which improve usability but do not represent a fundamentally new capability.
Is this a big deal?
Get stories like this every morning
The day's AI news in plain English, in your inbox at 8:30 IST. Free.
More from Anthropic
Anthropic Upgrades Claude Code Projects to Run Parallel AI Workers Hands-FreeAnthropic · 14 Nov 2026Claude designed FPGA scaler board for 6th-gen consolesAnthropic · 13 Nov 2026I asked Claude Fable 5.2 to turn my site into a 90s Silicon Graphics workstationAnthropic · 13 Nov 2026Anthropic is endorsing SB 53Anthropic · 13 Nov 2026Understanding and addressing AI harmsAnthropic · 13 Nov 2026This page was collected from platform.claude.com and summarised automatically by AIPulsee's pipeline. See how it runs.