Terminal Bench v4 vs Frontier Code v1.1?
aa has this new Frontier Code v1.1 bench which measures Agentic Software Engineering. I see that Kimi K3 does very well in this new bench while performing badly in Terminal Bench v4. What are the differences between these two benches?