Details
- Perplexity announces that the Claude Fable 5.1 model is now available inside Perplexity Computer for Pro and Max subscribers.
- Claude Fable is Anthropic’s long-horizon, task-oriented model designed for complex workflows, and 5.1 is an upgraded version over Fable 5.
- In Perplexity’s August WANDR benchmark of wide-and-deep research agents, Fable 5.1 ranked first with a soft F1 score of 0.601.
- The evaluation reports an average cost of $12.76 per task for Fable 5.1, delivering a 21% higher score and 37% lower cost than the earlier Fable 5 configuration.
- WANDR is Perplexity’s open benchmark of 500 real-world research tasks, used to compare production AI research agents on accuracy, completeness, and cost.
- Making Fable 5.1 available in Computer aligns Perplexity’s orchestrator stack with one of the top-performing long-context models on WANDR.
- The change is limited to paying tiers, signaling that Perplexity sees premium value in pairing its agent framework with Anthropic’s latest Fable iteration.
Impact
By integrating Claude Fable 5.1 into Perplexity Computer, Perplexity strengthens its position in the AI research-agent market, combining a top-ranked WANDR performer with its existing search and orchestration stack. This narrows the gap with rivals like OpenAI and Google that similarly pair strong foundation models with agent frameworks, while emphasizing cost-efficient performance on complex knowledge work.
