Anthropic Launches Opus 5.5: Reduced Prices and Enhanced Fable-Level Performance

Anthropic has unveiled its latest AI model, Opus 5.5, which was launched on Tuesday, setting new benchmarks in both coding and knowledge-based tasks, as reported by the company. This latest version reportedly surpasses the previous Fable model in numerous performance evaluations and has successfully tackled various informal assignments that Fable could not accomplish.

Notably, Opus 5.5 comes at a reduced cost compared to its predecessor. The new model will charge $20 for every million output tokens, a significant drop from the $25 fee associated with the previous version. Other pricing metrics have also seen similar decreases, and the efficiency has improved, resulting in quicker processing times.

The latest iteration is designed to enhance communication clarity, with Opus 5.5 less prone to using jargon and more inclined to present vital information upfront in its responses.

This release follows closely on the heels of the Opus 5, which debuted on July 24. The company has indicated that additional models, including Sonnet 5.5 and Haiku 5.5, are slated for launch within the forthcoming weeks, promising similar enhancements in performance.

Anthropic has stated that Opus 5.5 matches the Mythos model in terms of biological and cybersecurity features. Consequently, it is subject to the same safety protocols implemented for the Fable model. These measures restrict the model’s potential misuse, particularly in discovering vulnerabilities in compiled software or creating identifiable biological weapons.

This marks the first model release under the leadership of CEO Dario Amodei, who has advocated for a measured approach to AI advancements, aiming to synchronize progress in capabilities with alignment measures.

In a recent blog post, Amodei expressed, “I have become convinced that fully addressing the risks requires even more prudence… not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up.”

The safety protocols for Opus 5.5 resemble those of its predecessors, involving alignment testing and evaluations by external organizations, such as METR and Frontier Design. However, Anthropic has emphasized that they are already working on more advanced training and assessment systems for future models, including enhanced security and oversight mechanisms.

“As AI becomes more capable, public policy should play a larger role in ensuring that the systems people rely on are safe. That capacity takes time to build, and we’ve started to put the infrastructure in place to support it,” the company stated in their blog. They anticipate sharing more details about these initiatives soon.