The AI arms race shows no signs of slowing down this year, with Anthropic the latest major player to refresh its offering with the launch of Claude 3.
In short, Claude 3 is AI unicorn Anthropic’s new large language model (LLM), which means it’s now what powers the Claude AI chatbot. Technically, there are three different models that fall under the Claude 3 umbrella: premium Claude 3 Opus, second-tier Claude 3 Sonnet, and the unreleased Claude 3 Haiku.
The company not only claims that Claude 3 beats it old Claude 2.1 model, but boasts that it puts ChatGPT’s rival GPT-4 model and Google’s Gemini Ultra to shame. What is Claude 3 and how does it really stack up against the likes of ChatGPT and GPT-4? Here’s what we know so far.
What is Claude 3? Opus, Sonnet and Haiku Explained
Claude 3 is the new family of LLMs from Anthropic, an AI startup that’s been backed by (among others) Amazon to the tune of a cool $4 billion. The company is loosely part of the effective altruism movement, which in practice means it’s trying to develop artificial intelligence technology in a safe, responsible way that benefits the public, not just shareholders.
The Claude 3 moniker covers three new LLMs, the most capable of which is Claude 3 Opus. It’s Anthropic’s new premium AI model, available only to Claude Pro users. According to the company’s benchmarks, its graduate-level reasoning capabilities scored 50.4% when tested compared to GPT-4’s 35.7% score.
Below that sits Claude 3 Sonnet, which is what’s available if you don’t want to pay for a Claude Pro subscription. It’s still said to be highly capable, of course, with its graduate test score being 40.4% (again, compared to GPT-4’s 35.7% grade).
🔎 Want to browse the web privately? 🌎 Or appear as if you're in another country?
Get a huge 86% off Surfshark with this special tech.co offer.
Lastly, there’s Claude 3 Haiku, which is the currently unreleased third model. It sits at the bottom of the family tree in terms of sophistication and most capabilities, but is designed rather to be the fastest version and provide near-instant responses.
Haiku seems almost exclusively aimed as businesses as opposed to everyday users. Its speed means it’s the model that will help power things like customer-facing AI chatbots for websites, the ones that help you learn more about what you’re trying to buy, book appointment, and resolve other queries or complaints.
It’s also likely to be preferred by businesses who need to digest large amounts of data and reports in a time-sensitive manner.
Claude 3 Pricing: How Much Does Claude 3 Opus Cost?
As we’ve said, some Claude 3 models require you to pay for a Claude Pro subscription. Specifically, that’s the case with Claude 3 Opus, the most advanced version.
This effectively means that Claude 3 Opus pricing is set at $20 a month for consumers. Businesses wanting to use it will pay using the token system common for custom deployments of AI chatbot technology, which for Opus is $15 per million input tokens and $75 per million output tokens.
Claude 3 Sonnet is free to use on the open version of Claude AI that’s available to users all over the world, the “hard working” model in Anthropic’s own words, and companies can make it their own for $3 per million input tokens and $15 per million output tokens.
Right now, Claude 3 pricing for Haiku is therefore only available on a token basis: $0.25 per million input tokens and $1.25 per million output tokens.
Claude 3 vs GPT-4: First Look
We’re still testing Claude 3’s capabilities compared to key competitors like ChatGPT’s GPT-4 model. As such, we’re not yet in a position where we can share our thoughts.
That said, Anthropic has released a number of Claude 3 performance tests, which it says shows how the Claude 3 family stacks up against the likes of ChatGPT and Google’s Gemini Ultra LLM.
Of course, the key caveat here is that they’re Anthropic’s own tests and will be based, to a greater or lesser extent, on heavily engineered prompts designed to make the most of Claude 3’s capabilities.
That’s not to say Claude 3 isn’t hugely capable. It looks like it is. However, we can’t definitively say that it’s better or worse than ChatGPT, Gemini or any other chatbot until we’ve tested it fully ourselves.
Image Credit: Anthropic
How to Use Claude 3 Right Now
If you’re itching to see how Claude 3 responds to your favorite prompts, the good news is that middle model Claude 3 Sonnet has now rolled out to Claude AI in many countries around the world.
Better still, it couldn’t be easier to use Claude 3. Here’s how.
- Go to the Anthropic website
- Click the “Try Claude” button
- Login with your Google account (or create an account)
That’s it! You’ll now find yourself in the main Claude AI chat window, ready to start interacting with Anthropic’s helper. If you like what you see, you can then consider if upgrading to Opus and its additional capabilities is right for you.
