Anthropic releases Claude Fable 5.2 at a 20 percent lower price
Claude Fable 5.2 ranks first on three tests, but Artificial Analysis finds the highest cost for each task of all models it priced.
Anthropic, an AI company, released Claude Fable 5.2, which costs 20 percent less for each token than Fable 5.1. It also released Claude Mythos 5.2, a version with fewer biology and cybersecurity safeguards, only for organizations that Anthropic approves. The release came as the company began to sell shares to the public at a market value of $1.5 trillion.
How it works
Fable 5.2 and Mythos 5.2 are one model with different safeguards. In Fable 5.2, filters stop high-risk cybersecurity work, such as simulated attacks on a system. Anthropic says the filters stop about 55 percent fewer correct requests by mistake than in Fable 5.1.
The numbers
On the Intelligence Index of Artificial Analysis, Fable 5.2 scores 65 points, at $26.54 for each task. Gemini 4 Carbon, the next model, costs $5.48 for each task. In tests by Anthropic, Fable 5.2 did paid freelance projects as good as the work of persons, or better in 41.3 percent of the projects.
What is not known
Anthropic did not disclose the parameter count, the structure or the training data of the model. During training, Anthropic reduced two skills that speed up AI development. It says the model performs at about the level of Claude Opus 5.5 in them.
Insight: A lower price for each token is not a low cost for each task
Anthropic cut its price for each token by 20 percent. But Artificial Analysis prices Fable 5.2 at $26.54 for each task, against $5.48 for Gemini 4 Carbon. An earlier report said the share sale was near. The price cut and the top result seem to help the share sale: Anthropic shows the best score and rivals win on cost. Buyers must compare cost for each task.
Sources
Posted