Still, basic debuging and trouble shooting is where LLM generally shine. Any model can bisect git history and isolate newly introduced bug.

If model can not automatically reproduce bug, while human manually can... you got a problem in CI.

> Luna Max performing better than Fable 5.1 High

Perhaps you are reading too many benchmarks.

Edit for answer :

I agree Luna is great cheap model.

But if Fable was hitting security limits, yet was still included in benchmarks...

What flies better? Elephant or paper plane. You can make objective benchmark about that. But not much value for logistics company