AI company Anthropic is previewing its new “family” of Claude 3 models it claims can outperform Google’s Gemini and OpenAI’s ChatGPT across multiple benchmarks.
This group consists of three AIs with varying degrees of “capability”. You have Claude 3 Haiku down at the bottom, followed by Claude 3 Sonnet, and then there’s Claude 3 Opus as the top dog. Anthropic claims the trio delivers “powerful performance” across the board due to their multimodality, improved level of accuracy, better understanding of context, and speed. What’s also notable about the trio is they’ll be more willing to answer tough questions.
Anthropic explains older versions of Claude would sometimes refuse to answer prompts that pushed the boundaries of the safety guardrails. Now, the Claude 3 family will have a more nuanced approach with its responses allowing them to answer those tricky questions.
Despite the all-around performance boost, much of the announcement is focused on Opus as being the best in all of these areas. They go so far as to say the model “exhibits near-human levels of comprehension… [for] complex tasks”.
Specialized AIs
To test it, Anthropic put Opus through a “Needle In a Haystack” or NIAH evaluation to see how well it’s able to recall data. As it turns out, it’s pretty good since the AI could remember information with almost perfect detail. The company goes on to claim that Opus is quite the smart cookie able to solve math problems, generate computer code, and display better reasoning than GPT-4.
The technology isn’t without its quirks. Even though Anthropic states their AIs have improved accuracy, there is still the problem of hallucinations. The responses the models churn out may contain wrong information, although they are greatly reduced compared to Claude 2.1. Plus, Opus is a little slow when it comes to answering a question with speeds comparable to Claude 2.
Of course, this isn’t to say Haiku or Sonnet are lesser than Opus as they have specific use cases. Haiku, for example, is great at giving quick replies and grabbing information “from unstructured data”. Also, it’s not as good at answering math questions as Opus. Sonnet is a larger-scale model meant to help people save time at menial tasks and even parse lines of “text from images”, while Opus is ideal for large-scale operations.
Changing the internet
Both Sonnet and Opus are currently available for purchase although there is a free version of Claude on the company website. A launch date was not given for Haiku, but Anthropic states it’ll be released soon.
As you can probably guess, the Claude 3 trio is meant more for businesses looking to automate certain workloads. Your experience with the group will likely come in the form of an online chatbot. Amazon recently announced it’s going to be implementing Anthropic’s new AIs into AWS (Amazon Web Services) giving websites on the platform a way to create a customized Claude 3 model to suit the needs of brands and their customers.
If you’re looking for a model suited for everyday use, check out TechRadar’s list of the best AI content generators for 2024.