Details
- Perplexity announces that Grok 4.5 is now available as an orchestrator model inside Computer for Consumer Pro and Max subscribers, expanding the range of frontier models accessible through its platform.
- The team evaluated Grok 4.5 against five other orchestrator configurations using WANDR, Perplexity's internal benchmark focused on agentic research capabilities.
- In these WANDR tests, Grok 4.5 achieved the highest score among all evaluated orchestrator setups, indicating stronger autonomous research performance in the Computer harness.
- Perplexity reports that this performance advantage comes at roughly half the effective cost of using Claude Opus 4.8 as an orchestrator, positioning Grok 4.5 as a more cost-efficient option for power users.
- Grok 4.5 has been characterized in external benchmarks as a near-frontier, token-efficient model aimed at agentic tasks and coding workflows, so its addition to Computer Pro and Max is targeted at intensive research and automation use cases.
- Making Grok 4.5 available in the orchestrator slot suggests Perplexity is actively tuning multi-model routing for subscribers, balancing quality and cost across different frontier models.
- The announcement underscores that these results are based on Perplexity's own WANDR benchmark, rather than a public leaderboard, highlighting internal validation of Grok 4.5’s suitability for research orchestration.
Impact
By integrating Grok 4.5 as an orchestrator for Consumer Pro and Max, Perplexity strengthens its competitive positioning in AI-native research tools, offering near-frontier agentic performance at substantially lower cost than Opus 4.8. This can shift high-intensity research workloads toward Grok 4.5 inside Computer, narrowing the gap with leading AI assistants that already mix multiple frontier models and making premium tiers more attractive for power users focused on autonomous research.
