Text-output models; recommended order uses recorded benchmarks first, then estimated cost.
Filters & usage budget
Performance —
Loading the live model catalog…
SIDE BY SIDE
Model PK · 0/3
Estimates use the usage above. Selections survive filter changes. A listed API route does not guarantee access for your region, account or current uptime.
Choose + PK on up to three model rows to start comparing.
SOURCES & METHODOLOGY
A traceable basis for every choice
Capability evaluation
Epoch AI public data syncs every six hours. ECI is a composite index; other metrics retain their original scales and are never merged into a score. Only unique model identities are matched. See details for tested versions, reasoning settings and sources.
OpenRouter rates and capabilities sync every 15 minutes; Models.dev specs and providers every six hours. Each scenario selects a relevant metric. Benchmark sorting uses that metric, then cost. Unmatched data stays empty.
This module supports selection and comparison. It does not run paid inference. Validate accuracy, latency, citations and tool success on real samples before launch.