The current extrapolation provided by Vals predicts that an evaluated AI model from Anthropic will achieve a 0.60 on this benchmark in August. 0.60 represents advancing substantially past selected references across five different categories of AI R&D research.

Resolves based on https://www.vals.ai/benchmarks/rsi_index. The model has to be released in some capacity (like how Mythos preview/Gemini 4 Carbon were initially "released") by the end of August, but the score does not necessarily need to be published by then. If the RSI Index is officially retired, this market will N/A. If there are any other uncertainties, it will be resolved by an unbiased SOTA AI model.
Resolves Yes if Anthropic achieves a 0.60 or higher on the overall benchmark. Resolves No otherwise

Vals RSI Index Leaderboard and Methodology | Vals AI

Explore the Vals RSI Index for autonomous AI research. Compare agents on LLM R&D tasks against published, human and frontier-model reference results.