We follow AI models that launch under codenames, test them on hard benchmarks, and work out who built them.
StealthMark tests planning, spatial reasoning and visual design. Two of its benchmarks you can play yourself.
Leaderboard and map: https://stealthmodels.com/stealthmark
Play Foresight and Latch: https://stealthmodels.com/play
Videos: https://www.youtube.com/channel/UC0GW5Z8EQzZZ3phIwCM26_A
Still in stealth: Iceman Contributor, Glyph Cluster. Waiting for: Qwen 4.
Unstealthed: Ox Alpha is GLM-5.3-Flash. Space Bunny Alpha is an early MiniMax M3.1 Flash. Pixel Canary is Callstack Apex. Owl Alpha is LongCat-2.0.