Gemini 3's 50% time horizon, per METR

134

1.5kṀ68k

Dec 1

<1.5h

12%

1.5h - 2h

20%

2h - 2.5h

28%

2.5h - 3h

21%

3h - 3.5h

3.5h - 4h

4h - 5h

1.2%

5h - 6h

0.5%

6h - 7h

0.5%

7h - 8h

0.3%

8h - 9h

0.3%

9h - 10h

0.3%

10h - 11h

0.2%

11h - 12h

0.3%

>=12h

This market will resolve to the highest 50% time horizon, as reported by METR, for any Gemini 3 model released within a month of the first Gemini 3 announcement.

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.

People are also trading

Gemini 3 flash release in November 2025?

4% chance

Gemini 3 Flash released by November 23, 2025?

3% chance

Will GPT-5.1 have a longer METR time horizon than Gemini 3?

40% chance

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

Gemini 3.0 Pro outperforms GPT-5 on METR 50% time horizon?

79% chance

Gemini 3's execution time-horizon?

1,830

How many will Gemini 3.0 achieve? [Read description]

2.53

Grok 5's 50% time horizon, per METR

Before 2026, will Gemini 3.0 exceed GPT-5 in Metr estimated time horizon?

76% chance

R2's 50% time horizon, per METR

13 Comments

129 Holders

1.3k Trades

Sort by:

How does this resolve if METR doesn't evaluate any Gemini 3 model which is released within a month?

@jim I think the “within a month” thing means any model of Gemini’s released within a month of the first announcement, not METR’s analysis

@bens yes, but it's not guaranteed that any Gemini models which meet this condition will be evaluated by METR.

Just reaffirming that i think this market is moronic

@jim You are going to lose alot of mana when the results come out

@Blocksterpen3 well, I have orders up so feel free to take my manas

you guys are criminally retarded

gemini 3.0 will make the best METR 50% time horizon go from 2h17min to 5h+, noting that gemini 2.5 pro, which was already very good, had a 50% time horizon of 39 minutes.

i wanna have what you're having jim

@Bayesian what I'm having is faith in METR-style time-horizon extrapolation. 5h is at the high end, but it's not <2% (nor <5% tbc)

@jim So even higher than 5%?

@HenryE yes

@jim Gemini 2.5 is 39 minutes, well below o3.

A straight line is about 80 minutes. You have to also bet that Google won't be staying behind on this metric which it historically has been.

There's another market that gives about 50% it will tie gpt-5, which is consistent with the 3 to 4 months behind theory, albeit with some catch-up relative to Gemini 2.5