Comment by LorenDB

Seems to be more or less on par with GPT-4o across many benchmarks: https://x.com/Ahmad_Al_Dahle/status/1865071436630778109