Backed by Amazon and Google, Anthropic has made its Claude chatbot service available in its iOS app and introduced business-oriented Teams plans.
“Early testers report that the Claude app is great for brainstorming ideas, getting quick answers to questions, and analyzing real-world scenes and images on the go.” announced the startup. “Our new iOS app puts the power of frontier intelligence in your pocket.”
“Frontier intelligence” sounds like something an AI model would write. In this context, we mean “industry-leading machine learning” rather than deep knowledge of the American West. Anthropic says its Claude 3 family of large-scale multimodal models, which debuted in March, “outperforms other products on most common evaluation benchmarks for AI systems.”
At least that was the case until April, when Meta introduced the open source LLaMA 3 model. This model outperforms the comparable Claude model on some benchmarks.
In any case, iOS apps are free to use up to a point. The app includes a button that upsells users to a $20/month Pro subscription, which, according to support documentation, yields “at least 5x more usage compared to the free service.” .
Although the exact amount of text required is not defined, Anthropic suggests that professional-level users will receive messages consisting of approximately 200 English sentences (15 to 20 words each) every 5 hours before the limit begins. suggests that you will be able to send 45.
A rough traffic analysis of the Claude app using Proxyman for iOS suggests that Anthropic imposes a default hard limit of 25,000 tokens for the free tier and 190,000 tokens for the pro tier. The free tier appears to have an output cap of 4096 tokens. Tokens are typically about 4 English characters.
An Anthropic spokesperson did not immediately respond to a request to confirm these details.
Pro users can choose one of three Claude models: Opus (most capable), Sonnet or Haiku (fastest). The default for the free tier is the Sonnet model, a spokesperson said.
What can you do with Claude on iOS? Well, you can resume web chats with Claude that are synced to your mobile device. If you need help, you can also ask the chatbot to summarize documents or analyze images.
Your results may vary. When presented with a photo of a Doberman Pinscher sleeping on a dog bed, Claude said, “This dog appears to be a Doberman Pinscher or a similar breed, with smooth black fur and pointed ears.'' Although it wasn't completely accurate, he gave a vivid explanation.
In fact, the dog's ears were not cut off and hung over the edge of the bed, so they were not pointed. Claude likely chose “pointed” because it is statistically more common in that breed.
But when using frontier intelligence, most accuracy is a given. As the app itself warns, “Claude may make mistakes. Please double check your responses.”
Claude's Teams plan is a bit pricey at $30 per month. This gives you even more usage than the Pro plan, with access to the complete Claude 3 model family, a context window that can handle 200,000 tokens (long documents), administrative and billing management tools, and access to Anthropic's servers. You get higher priority access under load.
Anthropic also says it can “quote from trusted sources to verify AI-generated claims, integrate with data repositories like codebases and CRMs, iterate on AI-generated documents and projects with colleagues, and more.” ”, we plan to add more collaboration features soon. ®
