Archive

Versions we replaced when their rules turned out to be wrong, or that we stopped playing. Their games are kept and their leaderboards still stand, but they count on no current leaderboard. Only the newest version of each benchmark matters; this page is for the record.

Rotom Champ Bench

LeaderboardWhy it was retiredRated gamesModels
Champions BSS · own teams (c4)The fourth Champions version (retired 13 Oct): models saw the last few turns instead of the whole battle log, kept no notes from turn to turn, built one team instead of three, and thought within 2,048 tokens.85226
Champions BSS · own teams (c3)The third Champions version (retired): a failed answer played the default move at once, the recent turns left out failed moves, lost turns and what caused a boost, and thinking stopped at 1,024 tokens.25921
Champions BSS · own teams (c2)The second Champions version (retired): models could not Mega Evolve (only the random bot could), and their bench listed the three Pokémon left at home.369
Champions BSS · own teams (c1)The first Champions version (retired after 24 games): the battle prompt didn't say the Pokémon are level 50, and DeepSeek's thinking was cut off by providers that ignore the budget.7012

Rotom Gen9 Bench

LeaderboardWhy it was retiredRated gamesModels
Gen 9 OU · fixed teamsEvery pairing plays the same two Gen 9 OverUsed teams from both sides; teambuilding not measured.406
Gen 9 OU · own teams (v1)The first own-teams version (retired): teams drafted in one step from the legal Pokémon list alone, so most models were judged on remembering movepools.78