KalshiTechOpenUpdated Jun 29, 2026

When will AI solve another Frontier Math Problem?

6 markets · 4mo

AI overview· Updated 6d ago

The market sees a solution to a FrontierMath problem as unlikely before September (34% Yes) but likely by October (55% Yes), plateauing near 75% by year-end, implying a specific expected catalyst around October that is not explained by any provided headlines.

📊 Base rate: As of mid-2026, no AI system has publicly solved a single problem from the FrontierMath benchmark since its release in 2024, providing a historical prior of 0% for any given month, making the current odds for later months relatively high.

What's driving it

• The steep increase in Yes probability from September (34%) to October (55%) suggests a specific expected catalyst around October, but no recent headlines explain this shift.
• The low probability before August (14% Yes) indicates that the market sees no imminent solution in the very near term, consistent with the benchmark's difficulty.
• The plateau around 72-76% for November through year-end suggests residual uncertainty that a solution may slip into 2027.

The bullish case

• Rapid progress in AI reasoning models (e.g., chain-of-thought, self-play) has dramatically improved performance on math benchmarks, making a FrontierMath solution plausible soon.
• Major labs (OpenAI, DeepMind, etc.) are known to be targeting hard math problems, and a breakthrough could be announced at any time.
• The market's 55%+ odds for October onward imply a consensus that a solution is more likely than not within that window.

The bearish case

• FrontierMath problems are designed to be extremely difficult, requiring deep mathematical insight that current AI systems still lack.
• No AI has solved any FrontierMath problem in the two years since its release, and the low odds before September (34%) reflect skepticism about near-term progress.
• The plateau below 80% for year-end suggests a significant chance that no solution will be found in 2026 at all.

What to watch

• No specific dated catalysts are identifiable from the provided data; the odds imply a catalyst around October but its nature is unknown.
• Any unexpected announcement of a FrontierMath solution from a major lab would immediately spike the relevant market to near 100%.
• A major AI conference or model release in late 2026 could serve as a venue for such an announcement, but no such event is confirmed in the provided headlines.

AI-generated · grounded in recent news + odds · informational only, not advice. Verify on the source platform.

✨ Ask about this event