Let's fix bugs before they cause a global crisis. Updated SWE-sweep leaderboard sees OpenAI and Anthropic models double the number of bugs found, while halving costs. Astra is the best model, but still finds less than 10% of all real-world bugs tested. Details in 🧵
3
7
5
56
19,147
On the Pareto frontier: Luna 5.6, Opus 5.5, Astra 6

Oct 22, 2026 · 2:14 PM UTC

1
4
337
Astra 6 significantly outperforms all other models and gets closer than any other model to fully solving all bugs for one repository. This plot shows the distribution of % bugs resolved for every repository.
2
3
214
It's quite interesting that Astra 6 is also cheaper than Sol 5.6. This is a combination of taking less turns and also consuming less input tokens (however note that it's different model generations, Sol 6 would have probably been cheaper)
1
2
140