Home DEVELOPER Anthropic AI is outdoing itself: better image recognition, better coding

Anthropic AI is outdoing itself: better image recognition, better coding

0


Anthropic’s AI assistant Cloud is now available in an enhanced version. Cloud 3.5 Sonnet replaces Cloud 3 Sonnet. Sonnet is the “medium” version of Cloud. The little brother Haiku is cheaper but not as “intelligent”, while Opus is more expensive to use but more “intelligent” – basically. Because now Cloud 3.5 Sonnet is being touted as cheaper and smarter than Cloud 3 Opus.

Advertisement


Faster, too: The new Sonnet should solve tasks in half the time compared to the old Creation. This is particularly attractive for chatbot operators who want to outsource customer support from knowledgeable humans to compute AI. The provider plans to release Cloud 3.5 later this year in the cheaper version Haiku and the more powerful version Opus. Cloud has also been available in Europe for about five weeks.

especially The provider is proud on improvements obtained in internal tests when updating the programming software, i.e. fixing errors or adding new functions to the given open source code. While Claude 3 Opus masters 38 percent of the tasks set, Claude 3.5 Sonnet succeeds in 64 percent of cases.

Anthropic also seeks to significantly improve image evaluation, especially when interpreting tables and graphics. In addition, the new AI assistant can better transcribe text in images that are not fully captured. In several other benchmark tests, Cloud 3.5 Sonnet outperforms its competitors GPT-4o and Gemini 1.5 by a few percentage points. These tests are standardized and not very meaningful for the actual handling of chatbots.

Software-architektur.tv: Why does agility fail? – Reactions

Cloud 3.5 Sonnet is available in a free version for internet browsers and as an iOS app, although with rate limits. For paid use, these limits are set five times higher. Like the previous version, it costs three US dollars per million input tokens and 15 dollars per million output tokens, with a 200,000 token reference window. This indicates how much the cloud can process at once.

“Our goal is to significantly improve the trade-off between understanding, speed, and cost every few months,” says Anthropic, which also counts Amazon.com and Google among its investors. At the same time, Anthropic is working on additional offerings to make AI even more useful.

Now new is a workspace called Artifact. In a separate window, it displays the content that the cloud has created at the user’s request. The content can then be edited and reused. Currently only from the respective user, in the future also from colleagues. In the future, the cloud may lose its memory discount: developers are researching a feature called Memory, with which the cloud remembers previous interactions with a specific user and their preferences; it aims to increase personalization and efficiency.




(DS)

Software Testing: Acceptance Test-Driven LLM Development

NO COMMENTS

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Exit mobile version