Capability
Reasoning, coding, multimodal support, tool use, and large-context behavior are weighted highest for premium model selection.
Trust layer
How the AI model score works
Scores are editorial decision indices, not official provider benchmarks. The goal is practical model selection: which model is most likely to work well for a specific job, budget, and deployment constraint.
Reasoning23%
Coding17%
Tools/context26%
Speed/value/control21%
Formula
Default overall weighting
Efficiency
Speed and value matter more for high-throughput products, routing layers, and user-facing chat interfaces.
Control
Open weights, local deployability, and enterprise control improve the control score even when raw hosted-model intelligence is lower.
Sources
Provider docs, public leaderboard references, pricing pages, and practical engineering fit are combined with clear editorial labeling.