The newest arrival in the GenViz stable is Claude Fable 5.2. Choose Anthropic → Fable 5.2 in the model picker to use it on your business data, with the same connections and analysis tools you already use in conversations.
It joins Opus 5.5, Sonnet 5.5, and our other models. Fable 5.2 is the one to pick when a question takes many steps to answer.
Built for harder questions
In its Fable 5.2 announcement, Anthropic says the model is for its most ambitious, open-ended work, and that early testers found it could take on “more ambiguous, data-heavy tasks than previous models.” Anthropic reports output up to 30% faster than Fable 5.1, at $8 per million input tokens and $40 per million output tokens.
Several of the announcement’s customer examples are analysis checks: regional figures that don’t add up to the reported total, a bank balance that doesn’t match the deposits, and SQL that runs without errors but double-counts summary rows. Those are the same mistakes you look for when you check an answer in GenViz.
These figures and examples come from Anthropic and its early testers. How long a GenViz turn takes also depends on your queries, your data, and how many analysis steps the question needs.
For quick lookups and everyday follow-ups, Opus 5.5 and Sonnet 5.5 are good places to start. Use Fable 5.2 when a question has several parts or the data is messy enough that you would check the work twice.
Give it a question with several parts
Try a question that usually takes a few rounds of back and forth:
Compare net revenue by region for the last two complete quarters. Exclude refunds and internal test accounts, check that the regional totals add up to the company total, and explain any gap.
Open the returned tables and inspect the queries, including the date ranges, the filters, and how the totals were combined. Then dig into the details:
For the region with the largest decline, break the change down by product line and customer segment. Show the rows behind the biggest contributor.
From there, ask for a chart of the trend, save the result as an extract to compare later, or use Python analysis for a calculation that needs more than SQL. Every step stays in the conversation, so you can trace each conclusion back to the data behind it.
Choose how much reasoning to use
Fable 5.2 starts at high reasoning effort in GenViz, matching Anthropic’s documented default. The picker offers low, medium, high, xhigh, and max. Anthropic suggests starting at high: settings above it help on some tasks but not on others, so save xhigh and max for questions where you have seen them make a difference.
Higher effort uses more tokens on each turn. Lower effort is often enough for a straightforward lookup, even on this model.
Try Fable 5.2 in your next conversation
Open the model picker and choose Anthropic → Fable 5.2. Availability follows your organization’s Anthropic access. On the Team plan, an administrator needs to configure an Anthropic API key that can use the model.
Fable 5.2 uses the same read-only database tools and connection controls as the other models in GenViz. Your prompts and the context used to answer them, including relevant schema information and query results, are sent to Anthropic. Anthropic requires 30-day data retention for Fable 5.2, so an Anthropic organization set up for zero data retention can’t use it unless Anthropic has approved that. Our privacy policy explains the services involved.
Start with a question whose answer you already know and check the result. Then ask the one that has taken your team an afternoon of spreadsheets.
Explore conversation analytics or download GenViz to get started.