Humanity improves Cloud and piques Samsung’s interest

Anthropic has released updated models Claude Fable 5.1 and Mythos 5.1, betting not only on improved performance, but also on lower usage costs. These new products come amid anticipation for the next generation of OpenAI models.

According to the Artificial Intelligence Analysis Index, Fable 5.1 scored 4 points higher than the previous version and outperformed the previous Claude model as well as GPT-5.6 Sol in most tests. Meanwhile, the cost of input and output tokens remains at $10 and $50 per million dollars respectively. The price for reading cache data has dropped from $1 per million tokens to $0.25, which may indicate an increase in the use of flash memory to store KV caches.

However, there is a caveat: although caching is cheaper, complex tasks are more expensive. The average cost of each job in the artificial intelligence analysis intelligence index of “Fable 5.1” is $3.69, while the average cost of a job in “Fable 5.0” is $2.34, an increase of approximately 58%. At the same time, judging from the available information, “Fable 5.1” and “Myth 5.1” are almost the same model, but the former is subject to stricter protection restrictions.

Anthropic received special attention through its partnership with Samsung’s LSI division. The company tests Claude Code for semiconductor design tasks: the model helps create verification environments, prepare test scenarios and find potential problems. However, experts are still examining the results.

Other publications on this topic