Changelog

When the site changed, and when each benchmark, model and score was posted here. The date is the posting date; a score's own date is shown next to the score on its leaderboard. Atom feed.

Nov 20267 entries

  • benchmarks.finance goes liveA directory of 41 finance benchmarks and 358 published scores for 228 models, each quoted with its source and date. Leaderboards, best by use case, the directory, model compare and the method page are all up.LeaderboardsDirectoryCompareSite
  • Moneda Bench listed: six benchmarksMoneda GroundBench, MarketBench, KnowledgeBench, CiteBench, AdviceBench and PortfolioBench are listed with their method, metrics and submission rules. First results are still to come.How Moneda Bench worksGroundBenchMarketBenchKnowledgeBenchCiteBenchAdviceBenchPortfolioBenchBenchmark
  • Serif display typeHeadings and numerals are set in a serif, in larger sizes that are easier to read.Site
  • Logo tiles for model makersModel cards, bar charts and the cost map show each maker's mark beside the model name.LeaderboardsSite
  • Filters and sort on score tablesEvery leaderboard table can be filtered and sorted by column.LeaderboardsSite
  • About pageWhat the site is, who makes it, why Moneda Bench exists and how to send a correction.AboutSite
  • Submissions API for agentsPOST /api/submissions takes corrections, model, benchmark and product submissions, and messages as JSON. GET the same URL for the schema.APIllms.txtSite