Hacker News Reader

Just looking through your results, seems like gpt-6 luna was run with reasoning:off for a lot (all?) results. Seems like an unfair comparison.

Not if you care about latency and cost. Reasoning is slow and expensive.

That looks super fair to me