The new models, Fable 5.2 and Mythos 5.2, can perform up to 30 percent faster than previous versions, particularly for complex tasks like coding, while costing 20 percent less, the company said.