Which company has the best AI model end of September?

Platform comparison

Market context

The Arena.ai Text Arena leaderboard will determine which company operates the highest-ranked large language model as of 30 September 2026. The resolution uses the "Overall" ranking from the Text Arena leaderboard with style control disabled, measuring raw model performance across diverse benchmarks rather than user preference or stylistic factors. This metric has become the de facto standard for comparing frontier model capabilities among researchers and industry practitioners.

Historical precedent suggests sustained leadership on standardised benchmarks correlates with substantial computational investment and iterative training refinement. OpenAI, Anthropic, Google DeepMind, and Meta have alternated top positions on various leaderboards since 2023, with leadership changes typically following major model releases rather than incremental updates. The current 89% implied probability for a YES resolution reflects confidence that a single company will maintain clear separation from competitors, though the leaderboard's sensitivity to evaluation methodology means marginal ranking differences occasionally produce unexpected outcomes when benchmarks are recalibrated.

Traders should monitor announced model releases through Q3 2026, particularly from OpenAI, Anthropic, and Google, as new model versions frequently shift leaderboard positions substantially. The timing of training completion and public release schedules will prove decisive; models released in August or early September could alter rankings before the 30 September checkpoint. Additionally, changes to Arena.ai's evaluation methodology or benchmark composition—whether announced or implemented—could shift which company's model ranks highest, making platform announcements from the Arena team critical intelligence for position management.

Methodology

Sports-specific comparison page for Which company has the best AI model end of September?. Polymarket's live quote (Polygon order book) plus platform attributes for the three reference venues. Sports markets reward liquidity — Polymarket and Betfair are materially deeper than Kalshi or Manifold.

Resolution & payout

Sports markets typically settle on official final-whistle plus league confirmation. Polymarket uses UMA Optimistic Oracle with a source URL per contract — usually official league data feeds or ESPN/Soccerway. Two-hour dispute window, then smart-contract payout in USDC.

UK Frequently Asked Questions

When do sports markets resolve?

Typically within hours of the official final whistle. Polymarket uses UMA Optimistic Oracle with source URLs pointing at official league data feeds — e.g. the Premier League for EPL markets, UEFA for Champions League, FIFA for World Cup.

Can I bet on individual matches?

Yes, Polymarket lists every major Premier League / Champions League / World Cup match as its own market. Liquidity varies — top matches like El Clásico or a semi-final often have six-figure pools, lower-league games closer to three-figure.

Is Polymarket legal in the UK?

Polymarket is accessible to UK traders but is not UKGC-licensed. It operates under US CFTC jurisdiction. UK residents face no domestic legal prohibition on using it, but UKGC consumer protections do not apply. For UKGC-regulated alternatives, Betfair Exchange and Smarkets offer similar prediction-style markets.

Do I pay tax on prediction market profits in the UK?

UKGC-licensed platform profits (Betfair, Smarkets) are typically tax-free gambling winnings for UK individuals. Polymarket profits involve USDC crypto transactions, which HMRC treats as Capital Gains Tax events. Keep full transaction records and report via Self Assessment if gains exceed £3,000 per tax year.

How do I deposit on Polymarket from the UK?

UK traders typically fund Polymarket via Coinbase UK, Kraken or Revolut — buying USDC with GBP and transferring to a Polygon-compatible wallet (MetaMask, Coinbase Wallet). Polymarket's onboarding walks you through the bridging process. Typical GBP-to-USDC conversion costs 0.5–1%.