Starting today, users in Europe can access Claude, Amazon-backed AI startup Anthropic’s proper ChatGPT contender.
There are a few company names that have become synonymous with the advent of generative AI. OpenAI may have had the early-mover advantage, with ChatGPT becoming close to a household name. However, others have also been busy honing their craft and algorithms.
When Anthropic first released its latest large language model (LLM) family Claude 3 to the public in March 2024, we at the TNW editorial team were naturally very excited to try it. Alas, we were thwarted in our attempts as we were faced with the message that the model was not yet accessible in our region.
However, today we finally got to try out the ChatGPT rival for ourselves. (The API has been available to developers for a couple of months.)
The main thing we can say after playing around for less than an hour is that it is fast. Really fast. And the initial impression is that it is more thorough in many of its responses, even when we compare it to the paid version of ChatGPT.
It also feels more… human? And that is not just because Claude begins your relationship with asking your name, and telling you it’s nice to meet you.
Whereas it is still fairly easy to see that a text has been generated by ChatGPT (yes, PR and freelance pitchers, we can tell), I am not sure I can say the same for Claude. Which is simultaneously impressive and disconcerting.
Claude Opus, Sonnet, and Haiku
The Claude family comes in three versions – Opus, Sonnet, and Haiku. Opus is Anthropic’s ‘most intelligent’ model. According to its developers, it outperforms its peers on most of the common evaluation benchmarks for AI systems.
Claude is available through the web-based Claude.ai, an iOS app, and via a team plan for businesses. As a free user, you have access to Sonnet, albeit with a daily use limit, whereas Opus becomes available with a Pro subscription (€18 per month + VAT).
Claude claims to have “strong levels of comprehension and fluency in French, German, Italian, Spanish, and other European languages.” However, when prompted, it tells me it does not speak Swedish fluently.
But it does inform me — in Swedish — that it can provide me with simpler responses in Swedish should I prefer, however it would be easier to discuss more complex topics in English. This illustrates one of the reasons why training LLMs on data from smaller languages is incredibly important for all regions to benefit equally from the technology.
Claude is not multimodal — meaning that it cannot generate images, like OpenAI’s DALL-E, which is included in the paid version of ChatGPT. However it can process and analyse visual input.
On the same day Anthropic released Claude on our side of the pond, OpenAI launched GPT-4o. The two companies have funding from different big tech backers (the first from Amazon, and the second from Microsoft). Beyond LLM performance rivalry, Anthropic is positioning itself as the company who emphasises safety in developing AI.
The San Francisco-based startup claims to prioritise research on societal, safety, and security risks. In the words of its founders: “We founded Anthropic because we believe the impact of AI might be comparable to that of the industrial and scientific revolutions, but we aren’t confident it will go well.”
Either way, Claude is here. Let the European chapter of the chatbot wars commence in earnest.