The Growing Pains of Frontier Models: When Leaderboards Stop Separating and What to Measure Next
For AI researchers and practitioners tracking frontier model progress, the paper reveals that leaderboards are becoming less informative and offers a method to diagnose capability trade-offs and identify next useful benchmarks.
The paper shows that frontier model capabilities (coding and reasoning) are positively correlated (r=+0.72) but cooperation varies by lab and over time, with SWE-bench saturating while HLE and instruction-following retain discriminatory spread, and provides a diagnostic framework and seven falsifiable predictions for the next 12 months.
Leaderboards rank frontier models on independent axes but do not reveal whether capabilities reinforce or trade off across releases -- and at the frontier, this interaction is the more informative signal. We decompose paired SWE-bench and GPQA Diamond scores into a population coupling trend and per-release residual ($h$-field) that diagnoses capability emphasis and identifies which measurement or stress test is most informative next. Across 34 models from 10 labs (2024--2026), capabilities cooperate ($r = +0.72$, $p < 10^{-6}$), but cooperation varies by lab and over time: DeepSeek reversed from reasoning-rich to coding-first ($h$: $+11.2 \to -4.7$, 15.9-pp swing); Google maintains consistent reasoning emphasis; Anthropic oscillates between coding excursions and recovery. Cooperation is not static -- it cascades. Six open-weight architectures confirm a second capability transition at 30--72B, and SWE-bench is now saturating while HLE and instruction-following retain discriminatory spread -- signaling the next axis rotation. We provide a three-level playbook (locate, diagnose, rotate), a per-lab measurement-priority table, and seven falsifiable predictions with timestamped criteria for the next 12 months of frontier releases. Per-lab coupling slopes vary $5\times$ (Google $1.15$ vs. DeepSeek $0.23$), quantifying how efficiently each recipe converts coding gains into reasoning. Five April 2026 releases confirm the diagnostic out of sample ($r$ rises from $+0.72$ to $+0.75$). An interactive dashboard provides phase classification with actionable recommendations, $h$-field diagnostics, per-lab coupling trajectories, ODE-based scaling predictions, benchmark rotation guidance, self-steering demo, and live tracking of all seven predictions: https://zehenlabs.com/cape/.