Which company has the best AI model end of September?

Platform comparison

Market context

The Arena.ai Text Leaderboard will determine which company owns the highest-ranked large language model on 30 September 2026. The ranking uses head-to-head comparative evaluations across reasoning, coding, writing and general knowledge tasks, with no style-control adjustments applied. Ownership matters: OpenAI, Anthropic, Google DeepMind, Meta and other developers compete for top position, and the market resolves to whichever firm's model sits first in the "Rank" column at the settlement checkpoint.

Historical leaderboard volatility offers context for the 89% implied probability. Through 2024 and early 2025, leadership rotated between OpenAI's GPT variants and Anthropic's Claude models as each released incremental improvements. No single company has held unchallenged dominance across a full calendar quarter. The current crowd confidence likely reflects OpenAI's recent release cadence and Claude's plateau in relative performance, though leaderboard rankings shift on weeks-to-months timescales when new model versions launch. Kalshi and Betafair users pricing this at similar levels; Polymarket's fee structure (2% taker) and decimal odds display (versus implied probability formats on competing venues) create minor arbitrage opportunities for cross-platform traders monitoring this category.

Traders should track scheduled model releases from major labs through Q3 2026. Anthropic, Google and Meta typically announce capability updates via technical blogs or conference presentations; OpenAI's release patterns remain less predictable. The Arena.ai leaderboard updates continuously as new evaluations complete, meaning late-August and early-September announcements could shift the September 30 snapshot materially. Regulatory changes affecting model availability in specific regions, or Arena.ai's own methodology updates, represent secondary risks to resolution certainty.

Methodology

This page compares Which company has the best AI model end of September? specifically across Polymarket, Kalshi, Betfair Exchange and Smarkets. The live probability is the Polymarket mid; the comparison columns summarise each venue's fee structure, KYC, settlement currency and payment rails. Every CTA routes to Polymarket Alternative, which mirrors the Polymarket order book at 0% fees.

Resolution & payout

Polymarket settles via UMA Optimistic Oracle on Polygon. A proposer posts the outcome with a bond, the two-hour window runs, then the smart contract pays USDC.

Kalshi settles USD through the CFTC-regulated clearinghouse — the cleanest variant, with heavier KYC. Betfair Exchange settles in account currency (GBP/EUR), net of 2-5% commission. Smarkets follows the same model as Betfair with a lower default 2% commission.

FAQ

What does Polymarket cost vs Kalshi?

Polymarket: 0% fees, only Polygon network costs (~$0.01/trade). Kalshi: up to 7% per trade plus spread. For high-frequency traders, Polymarket is dramatically cheaper.

Which platform has the deepest liquidity?

Polymarket — by a wide margin. Top markets reach $50-500M volume, Kalshi ~$200M cumulative, Betfair similar. Deeper liquidity means your trade moves the quote less.

What about Smarkets as an alternative?

Smarkets is a UK betting exchange with a lower default commission (2%) than Betfair. Liquidity on political markets is below Polymarket, comparable to Kalshi. Geo-blocked in many jurisdictions.

Which platform is accessible globally?

Polymarket is geo-blocked in the US/UK/EU. Kalshi is US-only. Betfair and Smarkets are UK-restricted. Polymarket Alternative has a different geo footprint and routes to Polymarket's order book at 0% fees.

Are all these platforms regulated?

No. Kalshi is CFTC-regulated (US). Betfair and Smarkets are UK Gambling Commission licensed. Polymarket operates without explicit regulation — a different risk profile than a regulated sportsbook.