Understand this story
On November 4, 2026, OpenRouter released a conversation on the New Defaults program, featuring OpenRouter's cclark and Mercor's OsvaldNitski in a discussion about evaluation. The two discussed how, against the backdrop of rising inference costs, teams can use evaluation to weigh accuracy, cost, and latency when choosing models, and how evaluation helps teams decide when to optimize, switch models, or build their own models. OpenRouter's main post only included a link to the full episode and did not disclose any further follow-up arrangements.
AI-generated from public reports
Latest development · Report
OpenRouter talks with Mercor about how to choose models through evaluationDevelopments
1 confirmed developments
OpenRouter's cclark and Mercor's OsvaldNitski discuss evaluation
Source:OpenRouter
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 4
OpenRouterFirst-party
OpenRouter talks with Mercor about how to choose models through evaluation
On OpenRouter's New Defaults program, @cclark and Mercor's @OsvaldNitski discussed how, as inference costs rise, teams can use evaluations to weigh accuracy, cost, and latency, and decide when to optimize, switch models, or train their own models. The main post only includes a link to the full episode.
OpenRouterFirst-party
OpenRouter talks with Mercor: Using evals to decide model trade-offs
On the New Defaults program, OpenRouter's cclark spoke with Mercor's OsvaldNitski about how to choose models among accuracy, cost, and latency as inference costs rise, and how evals help teams determine when to optimize, switch models, or build their own.