OWL Claude
OWL / FABLE 5.1

Check your Claude Fable 5.1 relay.

A dedicated Fable 5.1 test: 11 checks for thinking signature fields, tool use, PDF, streaming, protocol and token usage. Inspect the evidence behind every score.

Independent community tool · ads never change scores
Claude Fable 5.1Anthropic Messages API

Use the API endpoint, not the login page. HTTPS only; redirects are blocked.

1dayapi: log in, select the default group, then copy an existing sk-1d-… API key. A 1dk_ login key cannot call inference.

Only Fable 5.1 is supported. You may enter its dated, thinking or vendor-prefixed model name.

Thinking probe: adaptive · xhigh effort. No manual mode selection needed.

Tests use your API balance. Regular full runs make up to 25 short requests; the thinking request permits up to 16,000 output tokens. Long-context runs can add ~332k to ~1.482M input tokens. Actual token counts and prices depend on your provider.

Your key is sent over HTTPS to OWL’s Cloudflare backend and then only to the endpoint you enter. It is used in memory for this run, never written to reports or our database. Reports stay in this tab unless you save or publish them.

METHOD / 01

Understand the rules

Weighted mean of scored checks. ≥70 passes; 50–69.9 needs review; below 50 fails. Critical adapter findings cap the verdict. Skipped or network-blocked checks reduce coverage and cannot create a complete pass.

Adapted from Veridrop public probes for Fable 5.1 on Cloudflare. OWL ruleset 1.1.1. The live Veridrop deployment is newer than its public source; exact scores may differ.

Claude Fable 5.1
Evidence and sub-checksWeight
Identity reply5
Behavior patterns15
Thinking signature fields25
Model consistency10
Knowledge checks10
PDF recognition8
Structured output12
Protocol fields5
Stream integrity5
Token usage10
Message identifiers5
Long-context recall15

Source & implementation notes ↗ · Understand the rules ↗ · Veridrop · AGPL-3.0

Reports on this device

Only reports you choose to save appear here. Up to 10 reports; API keys are never saved. Filter by endpoint or model.

Public reports & comparison

Only reports voluntarily published by users appear here. Comparisons use complete reports from the last 30 days, grouped by endpoint, model, mode and context setting, with at least 3 samples. This is not a reliability or uptime ranking.

Sample comparisons