Which company has the best AI model end of September?
Anthropic is the favorite because it currently appears to have the strongest position on the exact type of arena-style leaderboard that will decide the market. I still leave meaningful room for a late move by OpenAI or Google, so my estimate is below the current market price.
Analysis
Anthropic enters the final stretch with real momentum rather than just name recognition. The most relevant current signals suggest Claude-based models are already near the top, and at least one arena-style ranking has Anthropic models edging out the rest by a very small margin. That matters because this market does not reward the strongest overall brand or the best composite benchmark profile; it rewards the company that owns the single top model on the exact Text Arena Overall leaderboard at a specific September snapshot.
The main reason to stay below an extreme probability is that the margin appears thin and the field is still highly competitive. Google and OpenAI both have frontier models that can improve quickly, and either company could plausibly jump ahead with a new release, a tuning update, or a scoring shift that lands before the check date. A leader with only a narrow Elo advantage is not secure when the market still has several weeks left and the source is a live leaderboard that can move on relatively small performance changes.
The market price around the mid-80s implies strong confidence that Anthropic will simply hold its lead, but that may overstate how locked-in the ranking is. I agree Anthropic is the most likely winner, especially because current arena-style signals favor it and because leaderboard momentum tends to persist for a while. Even so, the exact noon ET checkpoint, the possibility of a rival launch, and the chance of a tight score reversal make this closer to the high-70s than a near-certainty.
Arguments
For
- Anthropic currently has the clearest evidence of strength on arena-style rankings, which is the most relevant signal for this market.
- Claude models have been broadly competitive across the kinds of user preference and instruction-following tasks that tend to do well on leaderboard systems.
Against
- OpenAI and Google remain close enough in capability that a single strong release could move them ahead before the resolution date.
- The current advantage looks narrow, so Anthropic does not need to fall far for another company to take first place.
Key drivers
- Anthropic currently appears to hold the strongest position on the arena-style leaderboard that matters most for resolution.
- The company has multiple frontier models near the top, which increases the odds that at least one Claude model remains first through September.
- The market is resolving on a single leaderboard snapshot, so small score changes or a late release can fully reverse the outcome.
- Competitors such as Google and OpenAI have enough model depth to overtake Anthropic if they ship a stronger update before the check.
Risk factors
- A late September release from Google or OpenAI could quickly push a rival model above Anthropic.
- The leaderboard gap appears narrow enough that ordinary ranking noise or a small Elo swing could flip first place.
- The exact noon ET check creates timing risk if Anthropic briefly leads but then loses the top spot later the same day.
- Any changes in evaluation behavior, model naming, or leaderboard updates could disrupt the current ordering unexpectedly.
Scenarios
Best case
Anthropic keeps the top slot on the Text Arena Overall leaderboard through the September 30 check, helped by stable scores and no rival release that meaningfully disrupts the ranking.
Most likely
Anthropic remains the favorite and probably stays near the top, but the final outcome depends on whether a competitor can land a late enough improvement to edge it out.
Worst case
A strong late update from Google or OpenAI overtakes Claude on the leaderboard, or Anthropic slips just enough in rank to lose first place at the exact check time.
More from this day
- techPolymarketEnded
Grok 4.6 released by...?
AI82%MKT2%Edge+80Hidden GemGrok 4.6 looks likely to be released by the deadline, with the strongest signals pointing to an August 7 launch or public availability. The market price appears far too low relative to the reported timeline, though a small risk remains that the release slips or is not broadly accessible enough to qualify.
- PoliticsKalshi1y
2026: Trump's bad year?
AI71%MKT10%Edge+61Hidden GemI think there is a strong chance 2026 becomes a recognizable bear-case year for Trump because multiple legal and institutional fights are already lined up to produce visible setbacks. The market’s single-digit yes price looks too low given how many independent downside catalysts are active.
- FinancialsKalshi13y
Will OpenAI or Anthropic IPO first?
AI34%MKT84%Edge-50HypedI put OpenAI first at about 34%, with Anthropic more likely to reach the public market first given the faster reported timeline and earlier filing reports. The current market price looks too confident in OpenAI's lead relative to the timeline risk.