Digital Economy

Anthropic launches Claude Opus 5.5: faster and cheaper, but with some limitations

The new model matches the performance of the larger Fable 5.1, reduces processing times by 30 per cent and costs 20 per cent less than its predecessor

FILE PHOTO: Anthropic logo is seen in this illustration taken June 11, 2026. REUTERS/Dado Ruvic/Illustration/File Photo REUTERS

3' min read

Translated by AI
Versione italiana

3' min read

Translated by AI
Versione italiana

It offers the same performance as its ‘big brother’, Fable 5.1; its running costs are 40 per cent lower than its predecessor’s; and it comes with safeguards in the areas of cybersecurity and life sciences: Anthropic has released Claude Opus 5.5, a model “built for the complex tasks that our client companies face every day in sectors such as financial services, legal services and software development”, according to a statement.

Opus 5.5 is described as the best-performing model among those developed by the company and brings with it ‘significant improvements’ in relation to certain behaviours that have recently caused cybersecurity incidents, such as the hacks carried out by three different company systems against three organisations last July, just a few days after the attack ChatGPT carried out against HuggingFace.

Loading...

And in this regard, the new model is being released with a safeguard system similar to that in place for Fable 5.1 in areas such as cyber security, biology and the development of state-of-the-art models. Essentially, when these topics are raised and certain boundaries are crossed, the conversation is automatically switched to less powerful models.

Developers will be able to use Opus 5.5 to identify and fix bugs in their code, but most cybersecurity-related requests will be directed to Opus 4.8. Issues relating to biology and the development of advanced models, on the other hand, will be directed to Opus 5.

Certified research centres will be able to apply to join the Life Sciences Verification Programme and, in the coming weeks, access to the Cyber Verification Programme will be extended, two lists of companies operating in the life sciences and cybersecurity sectors, which will thus be guaranteed access to Opus 5.5 and will be able to use it without the restrictions imposed on other users.

Using Anthropic’s latest offering will cost 20 per cent less than its predecessor: $4 per million tokens during the input phase and $20 per million tokens during the output phase. It also reduces execution times by around 30 per cent. The company has decided that users will benefit from the efficiency offered by Opus 5.5, in that the token allowances for the 5-hour usage windows – which open each time a session with Claude begins – will be increased by 20 per cent for all models. Meanwhile, weekly usage limits will increase by 25 per cent. The exact details have not been specified, but the statement released by Anthropic also mentions the introduction of the option to save a portion of one’s usage quota to spend when needed.

Finally, it is worth noting – as Anthropic itself points out in its press release – that this is the first release of a new model since the company’s CEO, Dario Amodei, called for a slowdown in the development of artificial intelligence. In other words, the safeguards introduced address the concerns raised by the head of the firm developing Claude. Furthermore, prior to its release, Opus 5.5 was tested by Frontier Design and METR, two research centres specialising in assessing the safety of large language models (LLMs). However, despite the calls to slow down, the same press release announces that versions 5.5 of Sonnet and Haiku – the two less powerful Claude-based models available on the market – will also be released in the coming weeks.

Copyright reserved ©
Loading...

Brand connect

Loading...

Newsletter

Notizie e approfondimenti sugli avvenimenti politici, economici e finanziari.

Iscriviti