Skip to main content
MANIFOLD
When will SWE-bench be solved?
11
Ṁ100Ṁ594
2030
0.6%
2024
0.6%
2025
88%
2026
3%
2027
3%
2028
3%
2029
1.3%
later

Resolves to the first year in which https://www.swebench.com posts a 90% resolved score in the “verified” category. The current score is 38.80% by Amazon Q Developer Agent.

Market context
Get
Ṁ1,000
to start trading!
Sort by:
bought Ṁ5 YES

Relatively significant chance of never (which resolves at later). Frontier LLMs are just going to get more and more contaminated with this benchmark's test data - at some point authors might just stop tracking