Skip to main content
MANIFOLD
OpenAI Astra Prop Bets
2
แน€250แน€64
Oct 2
50%
Scores at least 67 on Artificial Analysis Intelligence Index
50%
Costs at least $50 per million output tokens on release
50%
Scores at least 50% on Gene Bench Pro
50%
Resolves a FrontierMath Open Problems "Solid Result" Within One Month of Release
50%
Scores at least 80% on DeepSWE
50%
Scores at least 60% on Exploit Gym
50%
Scores at least 40% on Terminal Bench Science
50%
Scores at least 55% on Terminal Bench 4
29%
OpenAI claims it is AGI within one week of release
59%
It performs worse than GPT-5.6 Sol on at least one benchmark OpenAI reports in their release blog post or system card

All benchmarks (except for Artificial Analysis and Frontier Math Open Problems) will resolve according to what OpenAI claims Astra scores on them. In Lieu of that, I will look to third-party benchmarking, prioritizing the benchmark's official leaderboard (like https://www.terminal-bench-science.ai/ for terminal bench science)

Market context
Get
แน€1,000
to start trading!