78 AI models, 72 with a verification score, tested on the same 80 questions through the OpenRouter API. Every number below comes from that run; nothing is taken from vendor announcements.
Claude Opus 5.5 by Anthropic had the highest verification score, 85.0% (68/80); the median of 72 scored models was 62.5%.
Among models at or above the median score, the cheapest correct answer came from GLM 5.3 Flash at $0.00022; the most expensive, Grok 4.7, cost 145× as much per correct answer.
Reasoning tokens made up 86% of the output of the 58 models that think before answering. Their median cost per question was $0.0027, against $0.00045 for the 14 that don't.
Nemotron 3.5 Super VL was the fastest model at a median 715 tokens per second.
Cost per correct answer
Price per token says little on its own: a model that thinks at length or gets more questions wrong costs more per useful answer. Cost per correct answer is the billed cost per question divided by the share answered correctly. Shown for the 10 cheapest of the models scoring at or above the median (62.5%). The most expensive in that group, Grok 4.7, cost $0.031 per correct answer.
58 of the scored models think before answering, and those reasoning tokens are billed as output. They made up 86% of the output tokens of those models. Median score: 66.2% for reasoning models, 41.9% for the 14 others. Median cost per question: $0.0027 against $0.00045. Below, the models that spent the most reasoning tokens per question; all were run at reasoning effort low.
Median output tokens per second after the first token, over 3 streamed runs of the same prompt. Speed depends on the host OpenRouter routes to, so treat close values as equal.
This report uses the run of 2026-10-27, the last of October 2026. Scores come from 80 multiple-choice questions from SuperGPQA and MMLU-Pro. With that many questions, a score has a margin of error of several points, so small differences are noise. How every number is measured is on the methodology page; the current data is available as CSV and JSON, and the live rankings are on the leaderboard.
