Anthropic has just released Claude 3.5 Sonnet, the first model in their new Claude 3.5 family. This AI model sets new standards for intelligence in the industry, beating competitor models and even surpassing Claude 3 Opus in many tests. What makes it special is that it offers this high level of performance at the speed and cost of a mid-tier model. And the performance gains are noticable.
Claude 3.5 Sonnet is now free to use on Claude.ai and the Claude iOS app. People who pay for Claude Pro and Team plans can use it even more, with higher limits. Big companies can also use it through services like Amazon Bedrock and Google Cloud’s Vertex AI.
For businesses, Claude 3.5 Sonnet costs $3 for every million tokens it reads (input) and $15 for every million tokens it writes (output). It can handle conversations up to 200,000 tokens long, which is a lot of text.
Impressive performance
Claude 3.5 Sonnet does really well on tests that check how smart AI models are. It’s great at graduate-level thinking, college-level knowledge, and writing computer code. The model is better at understanding subtle meanings, jokes, and complex instructions. It’s also really good at writing high-quality content that sounds natural and relatable.
One of the best things about Claude 3.5 Sonnet is that it’s twice as fast as Claude 3 Opus. This speed, combined with its lower cost, makes it perfect for tasks like giving detailed customer support or managing complicated workflows with many steps.
In a special test where the AI had to fix bugs or add new features to open-source code, Claude 3.5 Sonnet solved 64% of the problems. This was much better than Claude 3 Opus, which only solved 38%. When given the right tools and instructions, Claude 3.5 Sonnet can write, edit, and run code on its own, showing advanced reasoning and problem-solving skills. It’s also great at translating code from one programming language to another, which is really useful for updating old software or moving code to new systems.
Visual intelligence
Claude 3.5 Sonnet is also very good at understanding images. It’s better than Claude 3 Opus at standard vision tests. This improvement is most noticeable when the AI needs to understand things like charts and graphs. Claude 3.5 Sonnet can also accurately read text from imperfect images, which is very useful in fields like retail, logistics, and finance, where AI might learn more from looking at an image or graph than from reading text.
New features: Artifacts
Anthropic is also introducing a new feature called Artifacts on Claude.ai. This lets users interact with Claude in new ways. When a user asks Claude to create things like code, documents, or website designs, these “Artifacts” appear in a special window next to the conversation. Users can see, change, and build on Claude’s creations in real-time, making it easier to use AI-generated content in their projects and work.
This new feature is just the start of making Claude more than just a chatbot. Soon, it will become a collaborative work environment. In the future, teams and even whole organizations will be able to safely store their knowledge, documents, and ongoing work in one shared space, with Claude acting like an on-demand team member.
Check out in the video below:
Safety and privacy
Even though Claude 3.5 Sonnet is much smarter, Anthropic has tested it thoroughly to make sure it’s safe to use. They’ve trained it to avoid misuse and have had outside experts test its safety features. Anthropic even let the UK’s Artificial Intelligence Safety Institute test Claude 3.5 Sonnet before releasing it.
Anthropic is serious about protecting user privacy. They don’t use any user-submitted data to train their AI models unless the user specifically allows it. So far, they haven’t used any customer or user data to train their models at all. Which is for sure a good direction.
What’s coming next
Anthropic plans to keep improving their AI models every few months, making them smarter, faster, and cheaper to use. They’ll be releasing Claude 3.5 Haiku and Claude 3.5 Opus later this year to complete the Claude 3.5 family.
The company is also working on new features to support more business uses, including ways to connect Claude with other business software. They’re exploring ideas like giving Claude a memory, so it can remember a user’s preferences and past conversations, making the experience more personal and efficient.
Anthropic loves hearing from users and encourages people to give feedback on Claude 3.5 Sonnet directly in the product. This feedback helps them improve Claude and decide what new features to add next.
Conclusion
The arrival of Claude 3.5 Sonnet shows that large language models are still getting better. This new AI is smarter and faster than older models in many ways. It’s good at understanding complex ideas and images, while still being affordable to use. Anthropic plans to keep improving their AI models in the future. This tells us that large language models haven’t reached their limits. There’s still room for these AIs to grow and do more. Claude 3.5 Sonnet is a big step forward, proving that we’re not close to seeing the end of what these models can do.