This market will resolve to the first 50% time horizon, as reported by METR, of Gemini 3 Pro. If Gemini 3 Pro Preview is evaluated first, the market resolves to Gemini 3 Pro Preview's 50% time horizon. If Gemini 3 Pro GA gets evaluated first, this evaluation determines the market's resolution.
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.
See also:
/Bayesian/gpt-52-pro-metr-time-horizon
/Bayesian/gemini-3s-50-time-horizon-per-metr (Pro Preview version of this market)
/Bayesian/gemini-3-pro-metr-50-time-horizon (this market)
/Bayesian/claude-sonnet-46s-metr-50-time-hori
/Bayesian/claude-sonnet-5-metr-50-time-horizo
/Bayesian/claude-opus-5-metr-50-time-horizon
/Bayesian/grok-420s-metr-50-time-horizon
/Bayesian/grok-5s-50-time-horizon-per-metr
1,000đ Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | áč3,382 | |
| 2 | áč1,589 | |
| 3 | áč126 | |
| 4 | áč71 | |
| 5 | áč47 |
Pessoas também estão operando
https://metr.org/assets/benchmark_results_1_1.yaml
gemini_3_pro: benchmark_name: METR-Horizon-v1.1 metrics: average_score: estimate: 0.709822 is_sota: true p50_horizon_length: ci_high: 444.530184 ci_low: 134.565523 estimate: 236.654674 p80_horizon_length: ci_high: 78.496574 ci_low: 21.399446 estimate: 43.435618 usage: usd: 0.0 working_time: 52126.13216666666 release_date: 2025-11-18
@MaxLennartson itâs so close to gpt5.1âs time horizon that it doesnt show up in the graph bc itâs covered completely
@Bayesian oh wow hahaha I was looking for it but hadn't been able to find the raw data link, I was too slow
@bens METR reported âabout 4 hoursâ which I interpreted as the same as reporting â4 hoursâ.
No other market in this category have you had to dig into the raw data of a yaml file.
@BraydonDymm I mean, you're welcome to wait until they update the little dot on their website, but it's a couple minutes short of 4 hours
yeah no other announcement has been this close to a full number for them to choose to round up in the announcement like this
@bens right, I understand thatâs what the raw data shows. I just had a different interpretation of what it means for METR to report the time.