Will any xAI Grok model score at least 30% on the FrontierMath Exam?

Analysis

The assessment of this forecast relies on the observed trajectory of AI performance on the FrontierMath benchmark, which has demonstrated rapid growth from approximately 2% in late 2024 to over 40% by early 2026. Historical data indicates that frontier models frequently achieve relative performance gains of 30% to 100% between major version releases. While current Grok iterations have scored in the 12-14% range, the broader base rate for advanced AI models suggests a high likelihood of reaching the 30% mark within a one-year timeframe. This analysis incorporates the Epoch AI leaderboard data, which tracks the progress of various models on these specific mathematical tiers. Consequently, the forecast reflects the consistent trend of rapid capability expansion in frontier-level language models.