BenchLeader

METR 50% Time Horizon leaderboard

As of 22 Nov 2026, the top score on METR 50% Time Horizon is 17 h 25 min by Claude Mythos Preview (Anthropic), ahead of Claude Opus 4.6 at 11 h 59 min, across 23 model configurations with a published result (METR, result dated 7 Apr 2026). How long a task can be, measured by the time it takes a skilled human, and still be completed by the model half the time. METR fits it from the model's results on its software and research tasks (suite version 1.1); a longer horizon is better. METR 50% Time Horizon is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 22 Nov 2026)

  • Claude Mythos Preview holds the top METR 50% Time Horizon score at 17 h 25 min (METR, 7 Apr 2026).
  • Claude Opus 4.6 is second on METR 50% Time Horizon at 11 h 59 min (METR, 5 Feb 2026).
  • METR 50% Time Horizon: How long a task can be, measured by the time it takes a skilled human, and still be completed by the model half the time. METR fits it from the model's results on its software and research tasks (suite version 1.1); a longer horizon is better (METR).
  • 23 model configurations have a published METR 50% Time Horizon result on BenchLeader (BenchLeader, 22 Nov 2026).
  • METR 50% Time Horizon does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

How long a task can be, measured by the time it takes a skilled human, and still be completed by the model half the time. METR fits it from the model's results on its software and research tasks (suite version 1.1); a longer horizon is better.

Published by
METRdata via Epoch AI Benchmarking Hub
Category
Agents & tools
Index weight
Reference only
Models
23
Data as of
22 Nov 2026
Newest result
7 Apr 2026

The publisher has posted no new results on this board since 7 Apr 2026. BenchLeader re-reads it every day, so newer models appear here as soon as they are published.

CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers.

Top 15
  1. 1Claude Mythos Preview17 h 25 min
  2. 2Claude Opus 4.611 h 59 min
  3. 3Gemini 3.1 Pro6 h 24 min
  4. 4GPT-5.2 (high)5 h 52 min
  5. 5GPT-5.3-Codex5 h 50 min
  6. 6GPT-5.4 (xhigh)5 h 42 min
  7. 7Claude Opus 4.54 h 53 min
  8. 8Gemini 3 Pro3 h 44 min
  9. 9GPT-5.1-Codex-Max3 h 44 min
  10. 10GPT-5 (high)3 h 23 min
  11. 11o32 h
  12. 12Claude Opus 4.11 h 40 min
  13. 13Claude Opus 41 h 40 min
  14. 14Claude 3.7 Sonnet1 h
  15. 15o1 (medium)39 min
23 of 23
#
1Claude Mythos Preview17 h 25 min–2026-04-07
2Claude Opus 4.611 h 59 min60.72026-02-05
3Gemini 3.1 Pro6 h 24 min59.52026-02-19
4GPT-5.2high5 h 52 min55.62025-12-11
5GPT-5.3-Codex5 h 50 min–2026-02-05
6GPT-5.4xhigh5 h 42 min63.02026-03-05
7Claude Opus 4.54 h 53 min56.92025-11-24
8Gemini 3 Pro3 h 44 min60.32025-11-18
9GPT-5.1-Codex-Max3 h 44 min–2025-11-19
10GPT-5high3 h 23 min58.22025-08-07
11o32 h60.82025-04-16
12Claude Opus 4.11 h 40 min51.72025-08-05
13Claude Opus 41 h 40 min51.72025-05-22
14Claude 3.7 Sonnet1 h49.02025-02-24
15o1medium39 min47.62024-12-17
16Claude 3.5 Sonnet21 min47.22024-10-22
17o1-preview20 min48.42024-09-12
18Claude 3.5 Sonnet 062011 min45.02024-06-20
19GPT-4o7.0 min39.42024-05-13
20GPT-4 Turbo Preview 11064.0 min45.42023-11-06
21GPT-44.0 min42.12023-03-14
22Claude 3 Opus4.0 min38.62024-02-29
23GPT-4 Turbo3.7 min40.92024-04-09

AI agent benchmark leaderboard · All LLM benchmarks

Cite as: BenchLeader, “METR 50% Time Horizon leaderboard”, data as of 22 Nov 2026, https://www.benchleader.com/benchmarks/metr_time_horizon_50.

METR 50% Time Horizon: questions

What does METR 50% Time Horizon measure?
How long a task can be, measured by the time it takes a skilled human, and still be completed by the model half the time. METR fits it from the model's results on its software and research tasks (suite version 1.1); a longer horizon is better. Scores are reported in lengths of time (how long the tasks take a skilled human), shown in minutes or hours; higher is better.
What is the top score on METR 50% Time Horizon?
As of 22 Nov 2026, the highest score on METR 50% Time Horizon is 17 h 25 min by Claude Mythos Preview (result dated 7 Apr 2026), ahead of Claude Opus 4.6 at 11 h 59 min. Scores come from METR and are re-read daily.
Which AI model leads METR 50% Time Horizon?
Claude Mythos Preview leads METR 50% Time Horizon with 17 h 25 min as of 22 Nov 2026 (result dated 7 Apr 2026), ahead of Claude Opus 4.6 at 11 h 59 min.
How many models have METR 50% Time Horizon results?
23 model configurations have a METR 50% Time Horizon result on BenchLeader, all taken from METR via Epoch AI Benchmarking Hub.
Who runs METR 50% Time Horizon and how often is it updated?
METR 50% Time Horizon is published by METR. BenchLeader re-reads the published results once a day and records the date each result was published.
Does METR 50% Time Horizon count toward the BenchLeader Index?
No. METR 50% Time Horizon is shown for reference but left out of the composite index.