Machine LearningAgents5 min reading time

Perplexity Cuts Agent API Research Costs and Adds Ultra Effort

Unite.AI
Read the full article
Perplexity lowered the cost per query for all Agent API effort levels while maintaining or improving research quality, with the Medium level improving accuracy and reducing costs by nearly 90%. They introduced an Ultra effort level that runs four agents in parallel for broader research, achieving the highest WANDR score at $21.71 per task.

More in Machine Learning