Since the release of Intelligence Index v4 in January 2026, Intelligence Index v5 has undergone its most significant benchmark restructuring. For the first time, this restructuring allocates 60% of the evaluation weight to private test sets or hidden answers, reducing the potential for contamination of publicly available question banks by training data and minimizing the likelihood of score inflation. The index still consists of nine evaluations. Within the 25% weight assigned to coding, the newly introduced FrontierCode 2.0 accounts for 15%, while Terminal-Bench 4.0 accounts for 10%. All tasks in FrontierCode 2.0 are derived from real open-source repositories, with the core scoring criterion being whether repository maintainers would be willing to merge submitted pull requests (PRs). This shifts the focus of coding ability evaluation from "runnability" to "mergeability." Among the released weighted models, DeepSeek V4.1-Flash 1017 (max) ranked first in the FrontierCode 2.0 category with a score of 28%, but its overall Intelligence Index v5 score was 32 points, tied with GLM-5.3 and trailing behind the top-ranked Claude Opus 5.5, which scored 58 points. The v5 version significantly widens the overall capability gap between open-source and closed-source frontier models, with the gap between the strongest open-source model and the top-ranked model expanding from 3 points in v4.3 to 12 points. The maximum overall score among released weighted models is 32 points. Additionally, v5 amplifies the performance gap between different thinking tiers, with notable differences in the same model's performance under cost-saving mode versus full-power mode. The cost per task for top-tier models can vary by more than 13-fold. The concurrently launched Progress Index v1 incorporates relevant methodologies, mapping major large language models since November 2022 onto a continuous ability scale to facilitate cross-temporal observation of model capability evolution. In this v5 evaluation, no single model ranked first across all categories, with individual strengths more dispersed than the attention given to overall rankings.