Pinned Tweet
Our model, RSO-F ranks joint 10th of all named teams on the board ForecastBench, among strong frontier models, and joint 35th on all 463 entries, in our first public benchmark.
Ahead of many impressive candidates including Claude Opus 5, Claude Fable, Google's NanoForecaster and many more.
As a small lab, we're happy to be in great company and look forward to improving on these results.
forecastbench.org/leaderboar…
Our models can produce increasingly accurate forecasts on questions where no public prediction market exists.
On Sep 27 our RSO-F model gave 80–85% odds that five top chess players would drop in the October FIDE rankings, while the median of 100 forecasters had these at ~50% or worse.
All 14 of our ranking calls turned out to be correct.
After a year in stealth: RysoAI
Forecasting and prediction that competes with frontier models.
ryso.ai
We believe forecasting models hold the key to a lot of breakthroughs and applications such as prediction markets, financial forecasting, robotics, AI agents and much much more.
Early performance of our models is proving to be comparative with or better than many frontier labs.
