Which company has the best Text Arena Math AI model end of September?
Anthropic looks like the leading contender and should remain favored, but the margin is not so large that a late-month swap can be ignored. I put the chance of Anthropic finishing first at 83%.
Analysis
Anthropic appears to be in the strongest position right now, with the latest available leaderboard snapshot showing a Claude model at or near the top of the Text Arena Math table. The market’s current price of 86.5% for Yes is broadly consistent with that, and the recent evidence suggests Anthropic has not just a temporary lead but a cluster of highly competitive models near the top of the ranking. When a company occupies multiple elite spots, it is usually harder for a rival to displace it quickly because small day-to-day score changes are less likely to push the top position away unless there is a clear new release or a major leaderboard reordering.
The main reason not to make this a near-certain Yes is that the market resolves on a specific end-of-month snapshot, not on average performance or recent dominance. Frontier leaderboard positions can move sharply when new model versions are introduced, scoring rules are updated, or enough user evaluations accumulate to shift Elo scores. The recent context already shows leadership changes in adjacent arenas and strong competitive pressure from OpenAI, which means Anthropic’s lead is real but not locked in. If OpenAI or another lab ships a materially stronger math-focused model before month end, or if an Anthropic score plateaus while a competitor gains momentum, the top ranking could flip on the exact check date.
Still, the burden of proof is on the challengers because Anthropic currently seems to have both the top rank and a broader presence near the top of the table. In markets like this, the most common failure mode for a strong leader is not a gradual collapse but a late surprise from a competitor that briefly overtakes it. That makes the No outcome plausible enough to matter, but not the base case. Overall, Anthropic remains the most likely company to hold first place at the end of September, though the probability should be a bit lower than the market-implied level because the snapshot-based resolution leaves room for a last-minute reversal.
Arguments
For
- Arguments for Yes: Recent leaderboard evidence points to Anthropic already occupying the top position in Text Arena Math.
- Arguments for Yes: Anthropic seems to have a strong cluster of high-ranking models, which increases the odds that one remains first.
Against
- Arguments against Yes: The resolution depends on one exact timestamp, so even a brief competitor spike can decide the market.
- Arguments against Yes: OpenAI and other frontier labs have shown enough momentum that a late-month flip is realistic.
Key drivers
- Anthropic currently appears to hold the top Text Arena Math position and multiple nearby top ranks.
- The market resolves on a single end-of-month snapshot, making late leaderboard volatility the main threat.
Risk factors
- A new OpenAI release or scoring surge could overtake Anthropic before the September 30 check.
- Leaderboard rank can shift quickly if underlying scores tighten or evaluation traffic changes.
Scenarios
Best case
Anthropic maintains the top Math leaderboard spot through the end-of-month check, with its leading model preserving a small but durable Elo advantage over the nearest rival.
Most likely
Anthropic stays near the top and probably finishes first, but the gap remains narrow enough that a late adjustment or rival improvement could still change the result.
Worst case
A competitor, most plausibly OpenAI, releases or benefits from a stronger model update that overtakes Anthropic right before the September 30 snapshot, pushing Anthropic to second place or lower.
More from this day
- CompaniesKalshi1y
Starbucks total global stores in 2026
AI84%MKT11%Edge+73Hidden GemStarbucks looks materially more likely than not to finish 2026 above 41,800 global stores. Current guidance and the existing store base imply a comfortable buffer unless execution deteriorates sharply.
- FinancialsKalshi13y
Will OpenAI or Anthropic IPO first?
AI34%MKT94%Edge-60HypedI think the market is heavily overconfident that OpenAI will IPO before Anthropic. Based on the latest reporting, Anthropic has the more imminent and better-advanced path to a public listing, so I put OpenAI-first at about 34%.
- PoliticsKalshi2y
Who will Trump pardon?
AI4%MKT50%Edge-46HypedI estimate Barron Trump’s chance of receiving a presidential pardon before January 21, 2029 at very low, around 4%. Absent a specific federal criminal exposure or public controversy, there is little reason to expect a pardon for a private family member.