WILDHAVEN BENCHMARK · REASONING MEDIUM

GPT-6 Astra

Openai · released 2026-09-03 · route gpt-6-astra · Model profile · Chess results · Sudoku results · Mines results · Nonograms results · FreeCell results · Starjar results · Nations Royale results · Hustle Sim results · Elemental Armory results · Micro Mouse results

Results

Score
10095%: 43 to 100 · 3 of 48 base raids so far
Stars per raid
3.00
Three-star raids / Townhearts down
3 / 3 of 3
Animals per party
3.67
Parties the game refused twice (default party taken)
0
Looks per raid
1.7
Landings the game refused
0 of 11 (0.0%)
Animals never landed
0
Retries for an unreadable or refused answer
0 (0.0% of requests)
Requests per raid
2.7
Model minutes per raid
0.8
Tokens per request (in / out)
12,113 / 205
Reasoning tokens per request
178
Raids on other rules (not counted)
0
API-equivalent cost per raid
$0.224
Seconds per request
18.3
Replies cut off at the length limit
0.0%
Aborted on our side (not counted)
0

By tier

TierRaidsDestructionStarsThree-star / TownheartGame's raiders
Hunt1100 (2 to 100)3.001 / 1100
Reckoning1100 (2 to 100)3.001 / 167
Outmatched1100 (2 to 100)3.001 / 166

Against the reference players, raid by raid

Its destruction minus each player's on the same raids, in points out of 100. Better and worse count the raids where it destroyed more or less.

PlayerRaidsGapBetter / worse
The game's raiders3+22.32 / 0
Everything at once3+16.32 / 0
Random landings3+32.33 / 0
Search, 64 tries3+2.31 / 0

Every raid