OpenAI’s GPT-5.6 series leads FrontierMath v2 Tier 4 leaderboards at 83% as of late August 2026, reflecting consistent progress on private research-level problems through iterative Python-enabled reasoning. Anthropic’s Claude models have closed ground with recent verified solves, including the Hadamard matrix of order 668 marked solved on August 12 after a researcher credited Claude alongside human collaborators, and earlier contributions to the 2-adic Galois group problem in July. OpenAI’s Astra internal model has also delivered batches of solutions to longstanding open questions, outpacing rivals in total attributed discoveries. Traders watch for the next Epoch AI-verified result on the expanding Open Problems set, where model releases, prompting scaffolds, and Lean formalizations can shift implied probabilities on which lab claims the next advance.
Experimental AI-generated summary referencing Polymarket data. This is not trading advice and plays no role in how this market resolves. · UpdatedSeptember 15
50%
September 30
47%
$120 Vol.
September 15
50%
September 30
47%
Any problems that Epoch AI already classified as solved before market issuance will not qualify for resolution. Only problems newly classified by Epoch AI as solved after market issuance may qualify. If Epoch AI adds new problems to FrontierMath: Open Problems after market creation, those problems will be considered for resolution.
This market will resolve solely based on Epoch AI’s classification or official determination that a qualifying problem has been solved, regardless of the degree of artificial intelligence involvement. A verifier-accepted submission, publicly posted solution, partial result, or third-party claim will not qualify on its own unless Epoch AI recognizes the problem as solved.
The primary resolution source will be official information from Epoch AI, including https://epoch.ai/frontiermath/open-problems.
Market Opened: Aug 28, 2026, 6:20 PM ET
Resolver
0x65070BE91...Any problems that Epoch AI already classified as solved before market issuance will not qualify for resolution. Only problems newly classified by Epoch AI as solved after market issuance may qualify. If Epoch AI adds new problems to FrontierMath: Open Problems after market creation, those problems will be considered for resolution.
This market will resolve solely based on Epoch AI’s classification or official determination that a qualifying problem has been solved, regardless of the degree of artificial intelligence involvement. A verifier-accepted submission, publicly posted solution, partial result, or third-party claim will not qualify on its own unless Epoch AI recognizes the problem as solved.
The primary resolution source will be official information from Epoch AI, including https://epoch.ai/frontiermath/open-problems.
Resolver
0x65070BE91...OpenAI’s GPT-5.6 series leads FrontierMath v2 Tier 4 leaderboards at 83% as of late August 2026, reflecting consistent progress on private research-level problems through iterative Python-enabled reasoning. Anthropic’s Claude models have closed ground with recent verified solves, including the Hadamard matrix of order 668 marked solved on August 12 after a researcher credited Claude alongside human collaborators, and earlier contributions to the 2-adic Galois group problem in July. OpenAI’s Astra internal model has also delivered batches of solutions to longstanding open questions, outpacing rivals in total attributed discoveries. Traders watch for the next Epoch AI-verified result on the expanding Open Problems set, where model releases, prompting scaffolds, and Lean formalizations can shift implied probabilities on which lab claims the next advance.
Experimental AI-generated summary referencing Polymarket data. This is not trading advice and plays no role in how this market resolves. · Updated



Beware of external links.
Beware of external links.
Frequently Asked Questions