Top SWE-Bench Pro score by Jan 1, 2027?

Ṁ1kṀ1.9k

2027

76.8 %

expected

ALL

40 - 49%

50 - 59%

18%

60 - 69%

30%

70 - 79%

24%

80 - 89%

17%

90 - 100%

This market predicts what the highest score on the SWE-Bench Pro public dataset leaderboard will be as of January 1, 2027.

Current top performers on SWE-Bench Pro public dataset (as of October 21 2025):

claude-4-5-Sonnet: 43.6%
claude-4-Sonnet: 42.7%

Resolution Criteria: This market will resolve to the score range that contains the highest score on the official SWE-Bench Pro public dataset leaderboard (https://scale.com/leaderboard/swe_bench_pro_public) on January 1, 2027.

Market context

Science

AI Benchmarks

Programming

Get

1,000

to start trading!

2 Comments

14 Holders

49 Trades

Sort by:

bought Ṁ15 NO

This market should be moving a lot after GPT-5.4 ended up improving ~2% on SWE-bench pro compared to GPT-5.2. That is a 3 month gap between models

The gap between Claude 4 Sonnet and Claude Opus 4.5 was half a year and in the time the score increase was 3.1%.

The gap between GPT-5.2 and GPT-5.3 codex, released 2 months apart, was 0.4% (per OpenAI Reporting). Given this data, how on Earth does this market think there is a 78% chance the score is going to increase at least 25% in the next 10 months?