Best AI model lab at the end of July
A near-term forecast for which company will hold the strongest overall AI model at month end.
Vibe Bench forecast tracker
Live market dataTrack release windows, benchmark thresholds, and which AI labs the market expects to lead. Every percentage comes from an exact active contract and refreshes throughout the day.
Source integrity
No invented odds
When an exact active market does not exist, we show no probability. Company forecasts stay separate from model-specific release forecasts.
Market leadership
Each outcome is its own Yes or No contract. Prices can move independently and may not add up to exactly 100 percent.
A near-term forecast for which company will hold the strongest overall AI model at month end.
A focused forecast for the company expected to lead the Text Arena coding field at month end.
A live forecast for the company expected to lead the Code Arena WebDev evaluation at month end.
A longer-range market forecast for which company will hold the strongest AI model at year end.
Upcoming AI releases
These are cumulative deadlines. A 70 percent Yes price by August means the market estimates a 70 percent chance of release on or before that checkpoint.
Cumulative estimates for the next public Claude Opus release. The final model name is not confirmed.
Cumulative estimates for the next public Gemini Pro model release.
Cumulative estimates for a future Mythos class release after the current Claude Mythos 5 model.
Benchmark thresholds
These contracts forecast public leaderboard thresholds. They are not measured benchmark results and remain outside the Vibe Coding Index.
Market estimates for the Text Arena score threshold reached by the next Claude Opus release.
Market estimates for the Text Arena score threshold reached by the next Gemini Pro release.
Complete model coverage
A direct market names the model family. A provider market only forecasts the lab and does not imply that every model from that company will lead.
How this tracker works
Every event uses a reviewed Polymarket slug. We do not search broadly and attach lookalike markets.
Public Gamma API responses are cached briefly and a six hour stale copy protects the page during a source outage.
Prediction prices never affect model scores, tentative placements, or official benchmark rankings.
Prediction markets reflect trader expectations and can be wrong. Contract rules determine settlement. Review each source market before using a price in a decision.