If you have the budget, gpt3/gpt3.5 are pretty great at it.

If you want to do it on your hardware/cloud instances.

You can go for gpt-j or flan-T5, they do great with basic prompts.

Or you can try a finetuned model, you might be able to get a finetuned T5 or there's BART and pegasus as well.

For summarisation even smaller models are pretty good so don't just jump to the biggest models. Compare the performance.