Accuracy you can check.

conare-embed-1 and conare-rerank-1 are the models Conare serves. Each score is our run of the benchmark's official data and answer key; LMEB is scored as its leaderboard scores it, beside the scores the leaderboard publishes for every other model, and is not a leaderboard submission.

Comparator scores are what each vendor, leaderboard or benchmark release published, or our run of open weights where marked. No competitor's search, embedding or reranking service was called. Where per-question results exist, gaps carry a 95% interval from a paired bootstrap over questions (10,000 draws). Some benchmarks were used to choose between candidate models during development; each one says so on its page. Nothing here is averaged across benchmarks.

On the two conditions benchmarks, the LLMs were given each document's first 2,000 characters, so only the searches whose deciding fact falls inside them are counted, for every model. Prices are list prices per 1M tokens, read on 13 October 2026; Conare's is its rate card of 15 October 2026.