Grok 4.5 Didn't Win a Single Benchmark — Then Undercut Opus by 17x on Cost Per Task

Editor's take

Grok 4.5's benchmark performance lagged behind competitors like Fable 5 and GPT-5.5, failing to secure any top rankings. This outcome underscores the ongoing challenge for new LLMs to immediately displace established leaders, even with significant investment.

The significance lies in the apparent trade-off between raw performance and economic viability. While Opus 4.8 demonstrably leads in benchmarks, Grok 4.5's drastically lower cost per task, reported as 17 times less, suggests a different strategic approach. This highlights a growing bifurcation in the LLM market: one focused on absolute capability, the other on accessible, cost-effective deployment for broader applications.

Future developments will hinge on whether Grok can bridge this performance gap without sacrificing its cost advantage, or if other models will adopt similar cost-optimization strategies. Observing whether benchmarks evolve to incorporate cost-effectiveness metrics, or if specific industries prioritize Grok's price point over marginal performance gains, will be crucial.