v0.8.7: Benchmark Data Freshness

• jsDelivr cache purging — GitHub Action now purges CDN cache after committing new data for faster propagation
• No disk cache — benchmark data fetched fresh from CDN on every launch for simplicity

v0.8.6: Cost Sorting & Open Weights

• Price sort columns — sort benchmarks by input, output, or blended price per million tokens via [s] cycle
• Per-model source detection — runtime matching of AA entries against models.dev data
• Source filter — [4] cycles through All / Open / Closed
• Region and type grouping — [5] and [6] toggle grouped layout with colored section headers

v0.8.5: Release Profile

• Optimized release binary — strip, LTO, single codegen unit, panic=abort (~6MB, down from ~11MB)

v0.8.0–0.8.4: Benchmarks Tab

• Dedicated Benchmarks tab — browse ~400 model entries from Artificial Analysis with quality, speed, and pricing data
• Creator sidebar with 40+ creators, classified by region and type with grouping toggles
• Quick-sort keys — [1] Intelligence, [2] Date, [3] Speed — press again to flip direction
• Dynamic columns, detail panel, TTFAT, AIME benchmarks and more

Other

• Provider categories — filter and group providers by type (Origin,...