Resolution criteria

• This market resolves to the organization or laboratory (e.g., OpenAI, Anthropic, Google, or others) that developed the model holding the highest general capability score on the official Epoch Capabilities Index (ECI) Leaderboard (interactive version at epoch.ai/eci) as of December 31, 2027.

• The resolution will be based on the general ECI, rather than domain-specific sub-indices (such as the Math ECI or SWE ECI).

• Highest score is determined by the model with the highest point estimate on the index. If there is an exact tie between models from different organizations, resolution will be split equally among the tying options.

• If Epoch AI has ceased updating or maintaining the ECI, or if the site is permanently down at the end of 2027, the market will resolve based on the last available snapshot of the index on the Internet Archive Wayback Machine from 2027. If no data exists from 2027 and the index is defunct, this market will resolve N/A.

Background

• Developed by the non-profit research group Epoch AI, the Epoch Capabilities Index (ECI) is a composite capability metric that aggregates performance across over 50 distinct AI benchmarks.

• Because individual benchmarks tend to saturate quickly, the ECI uses a statistical model to stitch together different evaluations, mapping them onto a unified general capability scale. The scale is anchored such that Claude 3.5 Sonnet is positioned at 130 and GPT-5 is positioned at 150.

• Major AI labs regularly contest the top position as they release new frontier models, with historical leaders including GPT-4, OpenAI o1, GPT-5.5 Pro, and Claude Fable 5.