Perplexity adds Claude Opus 5, at 57% less cost than Fable 5
By Steven Van ·
Opus 5 is now selectable in Perplexity and Perplexity Computer, ranking second only to Claude Fable 5 on Perplexity's WANDR agentic benchmark.
Perplexity has added Claude Opus 5 to both its answer engine and Perplexity Computer, its agentic workspace product. Anthropic's newest Opus model joins the roster of frontier models Perplexity subscribers can pick for search, research, and long-running agent tasks.
The WANDR numbers
Perplexity did not just flip the model on, it benchmarked it. On WANDR, Perplexity's evaluation for agentic web tasks, Claude Opus 5 was tested against six other models and outperformed all of them except Claude Fable 5, Anthropic's top-tier model. The kicker is cost: Perplexity reports Opus 5 delivered that second-place performance while being 57% cheaper to run.
Why the price-performance gap matters
For a product like Computer, which chains many model calls into a single task, per-call economics compound quickly. A model that gets close to frontier-best results at roughly half the cost is often the rational default for agent workloads, with the most expensive tier reserved for the hardest steps. That is the calculus Perplexity is signaling here, and it mirrors how Anthropic itself positions Opus 5 relative to Fable 5.
Where you can use it
Claude Opus 5 is available now as a selectable model in Perplexity and in Perplexity Computer. Model access on Perplexity varies by plan, with Pro and Max subscribers getting the widest frontier-model selection. This continues a steady drumbeat of Anthropic models on the platform, following Claude Fable 5's role as an orchestrator model in Perplexity earlier this month.
Sources: Perplexity on X.