AI Explained reconstructs an early comparison between GPT-5.6 Soul and Fable 5 from their system cards because Soul initially had only a limited release. The available evidence suggests that Fable 5 or its underlying Mythos model is slightly stronger overall, while Soul is priced at roughly half the API cost.
The benchmark details complicate any simple winner. Soul performs well on terminal tasks and uses far fewer tokens on an exploit benchmark, but Mythos appears more reliable at peak capability and leads on some health and cybersecurity evaluations. Trial counts, safeguards and test design also make direct comparisons uncertain.
AI Explained connects the release to a wider shift toward gated frontier-model access, government oversight and concerns about model distillation. The video also highlights Soul's weaker behavior around some sensitive requests and destructive actions, making its price advantage important but not sufficient for every deployment.
Watch on YouTube
