Grok 4.5 Tops Perplexity WANDR Benchmark
In a post on X, Perplexity CEO Aravind Srinivas shared that SpaceXAI's newly released Grok 4.5 model achieved the highest score on Perplexity's internal WANDR benchmark, which measures agentic research capabilities within a computer-use harness. Srinivas highlighted that Grok 4.5 achieves this top performance at half the price of Claude Opus 4.8 (high) and also outperforms Perplexity's own GLM 5.2 model.
The AI frontier is undergoing aggressive price-to-performance competition, with Grok 4.5 showcasing that highly agentic, complex search and reasoning models can be delivered at a fraction of the cost of previous leaders, pushing companies like Perplexity to re-evaluate their model serving choices.
* Grok 4.5 leads on Perplexity's WANDR benchmark, signaling superior agentic tool-use and search capabilities.
* Offered at half the price of Claude Opus 4.8, it sets a new baseline for cost efficiency in agentic knowledge work.
* Outperforming GLM 5.2 highlights its potential to disrupt the current model pipeline of major AI search services.
DISCOVERED
1h ago
2026-07-10
PUBLISHED
2h ago
2026-07-10
RELEVANCE
AUTHOR
AravSrinivas