GPT-5 Pro's 50% time horizon, per METR

This market will resolve to the highest 50% time horizon, as reported by METR, for GPT-5 Pro, as per their first evaluation if there are several. If they release a major update to GPT-5 Pro but keep the same name, this market will only consider the GPT-5 Pro version(s) that are close in capability to the version currently available in the OpenAI API, as of market creation (Oct 7 '25)

50% time horizon is a measure of AI autonomy based on the length of tasks that AI can do: roughly, it is the time that humans take to complete tasks that an AI system can successfully do 50% of the time. See METR's "Measuring AI Ability to Complete Long Tasks" for the technical definition. Claude 3.7 Sonnet, released in February 2025, was the leading model with a 50% horizon of 59 minutes.

Left bounds inclusive, right bounds exclusive.

Time horizon could vary based on the set of tasks used to measure it, so this market will be based on the time horizon for the most comprehensive set of tasks reported by METR (as of 2025, largely software and engineering tasks). This will be ambiguous if METR stops publishing time horizons across all of their autonomy tasks and only publishes separate results for different subsets; I might N/A in that scenario.

If GPT-5 Pro is not evaluated by METR in the next 2-3 months, I may N/A this market, or extend it if there is indication that this eval may be coming.

Related questions