5h agoLisanBench creator Lisan al Gaib says Opus 4.8 beats GPT-5.5 among non-thinking models but ranks fifth overallOpus 4.8 achieved 93.3% clean stops on the benchmark.SentimentSentimentPos100%Neg0%Users are excited about Opus 4.8's large performance gains on LisanBench when enabling thinking mode because it shows strong results compared to other models even without thinking.2 comments with sentiment. View comments.