Historical movement is visible. Current momentum is not.

Here, momentum means movement in a provider’s sampled frontier rating within the selected legacy artifacts. It does not mean market share, revenue growth, model-release pace, developer adoption, or current strength.

The archive contains 11 dated artifacts from May 8, 2023 through August 29, 2025. Methodology changed during that span.

The highest sampled provider value changes across the selected artifacts. That makes the archive useful for reconstructing competitive handoffs.

It does not support a stable rate-of-change comparison across every interval because the metric and evaluation population changed. Rank movement is historical description, not a forecast.

Meta begins at 827.0 on May 8, 2023, reaches a sampled peak of 1395.9 on May 19, 2025, and ends at 1288.0 on August 29, 2025.

A momentum account that stopped at the peak would be incomplete. The final decline remains part of the selected record, and third-party derivatives stay excluded because we leave ambiguous ownership out rather than guess.

• Beginning: 827.0 in May 2023.
• Sampled peak: 1395.9 in May 2025.
• Archive end: 1288.0 in August 2025.
• Rejected claims using 1339, +512, or -57 are not used.

The July 2026 current snapshot measures something different from the legacy line and covers a single date. We never append it to the legacy line or use it to calculate current lab momentum.

A future momentum product requires repeated reviewed snapshots under one stable method. Until then, Model Gauntlet publishes the historical archive and current table as separate evidence lanes.

1. The archive is eleven selected artifacts, not a complete daily or monthly series.
2. Methodology changed within the legacy window.
3. Provider coverage varies by artifact and missing observations remain missing.
4. This report makes no claim about present-day momentum or future performance.

See the methodology and corrections policy.