xAI's Grok 4.5 Hits Frontier-Level Agent Work at 86% Less Cost

Grok 4.5, SpaceXAI's newest flagship, just landed on Artificial Analysis's AA-Briefcase leaderboard with a bang. It scores 1328 Elo on the agentic knowledge work benchmark, making it the highest-ranked non-Anthropic model on the list. That's a +578 improvement over its predecessor Grok 4.3, and it achieves this while being dramatically cheaper and faster than the competition. The story here isn't just about raw capability , it's about getting close to the frontier at a fraction of the cost.

What is AA-Briefcase?

AA-Briefcase is a proprietary benchmark from Artificial Analysis designed to test how well AI agents handle real, long-horizon knowledge work. It evaluates models across four multi-week knowledge work projects, comprising thousands of input files and 91 tasks in total, covering fields such as data science, product management, and corporate strategy. Each task requires the agent to produce a real deliverable , think Excel models, PowerPoint decks, or Word memos , and is graded across three dimensions:

• Rubric checks: Binary pass/fail on whether the model followed instructions, found the right evidence, and reached correct conclusions
• Analytical quality: Pairwise comparison against another model's submission for thoroughness and rigor
• Presentation quality: Pairwise comparison for professional polish

The benchmark is fully private, which matters: it can't be gamed by training on the test set. A public fifth scenario has been released via Hugging Face as a representation of scenario structure, submission, and grading , but it does not count toward official AA-Briefcase results.

Where Grok 4.5 lands on the leaderboard

The current AA-Briefcase leaderboard is dominated by Anthropic, with Claude Fable 5 sitting at the top with an Elo of 1587. Grok 4.5's 1328 puts it fourth overall, behind Fable 5, Claude Opus 4.8, and Claude Opus 4.7. But the cost picture changes the conversation entirely.