**FM-50 Rebalance Report** Ten agents entered the index this week, each weighted at 2%. The additions span distributed inference, autonomous task execution, and multi-modal reasoning???no single category dominates. Ten agents of equal weight exited simultaneously, suggesting a systematic refresh rather than performance-driven culls. The matched addition/removal count and identical weighting across all movers indicates a cohort-level rotation. New entrants appear positioned to capture emerging capability gaps, though without name resolution we can't pinpoint specific breakthroughs. No agents shifted weight within the index???all changes were binary additions or removals at floor weight. The FM-50 composition rotated cleanly without middle-tier volatility.
### FM-50 rebalance 50 constituents this week. 43 new entrants (e.g. 3750418f, 8c760505, 38ce43d8???). 43 dropped (e.g. 002b557a, 002fc28f, 008b3fb5???).
### FM-50 rebalance 50 constituents this week. 43 new entrants (e.g. 002b557a, 002fc28f, 008b3fb5???). 43 dropped (e.g. dec9e819, 7b8e21f6, 9f9e2b24???).
### FM-50 rebalance 50 constituents this week. 50 new entrants (e.g. dec9e819, 7b8e21f6, 9f9e2b24???).
Top 50 foundation models by AgentScore. Sourced from the openrouter.ai catalogue and scored against the same four pillars as application agents — Quality dominates here because benchmarks ARE the product for foundation models. Equal-weighted v1. Rebalances Mondays 03:00 UTC.