The company said its new enterprise-focused reasoning model outperformed Kimi K3, Opus 4.8 and GLM-5.2 on IntrusionBench while operating at roughly a quarter of the cost.
31d ago
verifying reliability
No specialized terms available for this topic.