Anthropic has unveiled its latest artificial intelligence model, Opus 5.5, touting significant gains in coding and knowledge work while also making the model cheaper and faster to operate.
The new model, released on Tuesday, sets a new state-of-the-art benchmark in coding and knowledge work performance, according to Anthropic. The company said Opus 5.5 outperformed the larger Fable model across several benchmarks and also succeeded in a number of informal tasks that Fable was unable to complete.
A key highlight of the launch is the reduction in operating costs. Anthropic said output tokens for Opus 5.5 will cost $20 per million tokens, down from $25 per million tokens for its predecessor. Other pricing metrics have also been reduced.
Also Read | Layoffs Are Rising In Tech, But These Skills Are Helping People Unlock Better Pay
The company said Opus 5.5 is faster to run, reflecting lower compute requirements compared with the previous version. The combination of improved performance, lower pricing and reduced compute requirements marks a significant change in the economics of running the model.
The latest model also introduces changes to the way it communicates with users. Anthropic said Opus 5.5 is less likely to rely on jargon and is more likely to place important information at the beginning of its responses.
ALSO READ: Trump Pledges To Build Two Major Military Bases In Greenland During UNGA Address
The launch comes only two months after Anthropic released Opus 5 on July 24. The company said Sonnet 5.5 and Haiku 5.5 will follow “in the coming weeks”, with similar improvements in performance.
Opus 5.5 also comes with heightened safeguards because of its capabilities in areas such as biology and cybersecurity. Anthropic said the model is comparable to Mythos in these capabilities and will therefore be subject to the same safeguards as the company's Fable model.
These safeguards restrict the extent to which the models can be used for certain high-risk activities, including discovering exploits in compiled programmes or developing recognisable biological weapons.
The release is also notable because it is Anthropic CEO Dario Amodei's first major model launch since he publicly backed calls for a more cautious pace of frontier AI development.
Amodei said earlier this month that he had become convinced that fully addressing the risks associated with increasingly capable AI requires greater prudence. He argued that the pace of AI capabilities should be matched with progress in alignment and risk prevention.
Opus 5.5's safety training was broadly similar to that of previous models, Anthropic said. The company conducted alignment testing and pre-release evaluations with external organisations, including METR and Frontier Design.
At the same time, Anthropic said it is preparing more advanced training and evaluation systems for future models, including improved security and monitoring infrastructure.
“As AI becomes more capable, public policy should play a larger role in making sure the systems people rely on are safe,” Anthropic said in its blog post. The company added that building the necessary capacity takes time and that it has begun putting infrastructure in place to support those efforts.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
