
After the launch of Perplexity training model, the cost dropped to one third of Opus
Perplexity published a GLM 5.2-based post-training model, stating that it achieves close to Claude Opus 4.8 in proxy assignments, at a cost of about three-thirds of the latter
















