Informal comparison
Keln AI Gateway Evaluation
Compare AI gateway caching, routing, latency, and cost under repeated agent workloads.
Explore the original evidence
Explore the original evaluation, its methods, and reported findings.
Open evalCoverage
A vendor-run comparison of OpenRouter, Vercel, Ramp, and Keln using DeepSeek V4.1 Flash over roughly one day.
Methodology
The same recorded agent turns are replayed across gateways, with synchronized pauses and requested maximum reasoning. Measurements cover cache reuse, provider changes, time to first token, throughput, reported reasoning tokens, and usage-based cost. Keln evaluates its own service; raw responses and the detailed ledger are unpublished. A prepaid-wallet limit shortened one gateway run.