BRO… GLM 5.3 Flash (ox alpha) pushed the Pareto frontier. it matches Opus 4.8 on AA’s intelligence score but is ~45x cheaper: > 57 intelligence vs 60 for GLM-5.3 > $0.045/task vs $0.68 for GLM-5.3 > 3x lower attention compute + 4.4x smaller KV cache at 1M context Intelligence is getting cheaper brutally fast.