xAI's Grok 4.6 Matches GPT-5 Performance at Lower Price Point

Grok 4.6, a language model developed by xAI, achieved the same score as OpenAI's GPT-5.6 Sol in performance evaluation. On the artificial intelligence benchmark index published by third-party organization Artificial Analysis, both models scored 61 points, tying for second place behind Anthropic's Claude Opus 5. This result once again highlights the intensifying competition among top-tier models.

In recent years, performance competition in large language models has rapidly converged among top-tier models, making it difficult to differentiate based on simple accuracy metrics alone. In this context, two evaluation dimensions are gaining attention: "agent performance" and "cost." An agent refers to an operational mode of artificial intelligence that autonomously completes multiple steps following user instructions, and its importance in industrial applications is growing because it directly enables automation of complex tasks.

Looking at specific numbers in agent tasks, Grok 4.6 completes complex workflows in approximately 53 steps. Meanwhile, Claude Opus 5, which receives the highest rating in the same category, requires 103 steps for similar tasks. The ability to accomplish equivalent or superior processing in fewer steps leads to improved processing speed and reduced API costs, lowering barriers for enterprises when actually integrating these systems into their infrastructure.

Price differences are also significant. Grok 4.6's pricing is set more than 60% lower than Claude Opus 5, revealing a strategy to establish a competitive position against both OpenAI and Anthropic on both performance and price fronts. The ability to deliver high performance while maintaining low costs increases appeal to cost-conscious enterprise users and developers.

xAI is an artificial intelligence company founded by Elon Musk, and the Grok series continues to be developed in coordination with the X (formerly Twitter) platform. The ongoing updates to Grok by the company can be understood as driven by the intent to increase presence in the enterprise-grade artificial intelligence market where OpenAI and Anthropic currently lead. This evaluation result can be viewed as one achievement of such strategy.

What this result demonstrates is that while performance among top-tier models is converging, competition on practical aspects—specifically "how efficiently" and "how affordably" models can be used—is intensifying. In particular, the efficiency of agent functionality directly impacts operational costs for enterprises integrating artificial intelligence into business systems, and it is anticipated that factors like step count and unit cost—"usability" indicators—will increasingly influence model selection beyond benchmark scores alone. The trend of focus shifting from benchmark scores to real-world cost performance is expected to accelerate going forward.