Follow and star “AIGC Developer“
Delivered daily at 08:45
Recently, Anthropic released shocking news: the new Claude 3 family has arrived! It is said that the strongest version, Claude 3 Opus, has surpassed all large models, including GPT-4, in reasoning, mathematics, coding, multilingual understanding, and vision, demonstrating near-human understanding capabilities.
Claude 3 is a large language model (LLM) developed by AI unicorn Anthropic. Technically, Claude 3 consists of three different models: the advanced Claude 3 Opus, the second-tier Claude 3 Sonnet, and the yet-to-be-released Claude 3 Haiku.
The company claims that Claude 3 surpasses its old model Claude 2.1 and provides data showing that it outshines competitors like the GPT-4 model and Google’s Gemini Ultra. What exactly is Claude 3, and how does it compare to GPT-4? Let’s take a closer look.
What is Claude 3? Explanation of Opus, Sonnet, and Haiku Versions
Claude 3 is a family of large language models launched by Anthropic, an AI startup supported by investments totaling up to $4 billion, including backing from Amazon. The company is involved in the effective altruism movement, meaning it aims to develop AI technology in a safe and responsible manner to benefit the public, rather than just for profit.
Claude 3 includes three new language models, the most powerful being Claude 3 Opus. This is Anthropic’s latest advanced AI model, available only to Claude Pro users. According to the company’s benchmarks, its reasoning capabilities scored 50.4%, while GPT-4 only reached 35.7%.
Next is Claude 3 Sonnet, which you can use if you don’t want to pay for a Claude Pro subscription. Of course, it is still considered very capable, with a test score of 40.4% (also higher than GPT-4’s 35.7%).
Finally, there is Claude 3 Haiku, the third model that has not yet been released. It is the least complex and feature-rich of the three, but it is designed to be the fastest version, providing nearly instant responses.
Haiku seems to be specifically aimed at enterprises rather than general users. Its speed means it will help drive things like AI chatbots for website customer dialogues, which can assist you in understanding what you want to buy, making appointments, addressing other inquiries, or complaints. It may be favored by businesses that need to digest large amounts of data and reports in a timely manner.
Claude 3 Pricing: How Much Is Claude 3 Opus?
As we mentioned, some Claude 3 models require you to purchase a Claude Pro subscription. Specifically, this applies to Claude 3 Opus, which is the highest version. The pricing for Claude 3 Opus is $20 per month. Businesses wishing to use it will pay through a token system customized for common AI chatbot technologies, with a cost of $15 per million input tokens and $75 per million output tokens for the Opus model.
Claude 3 Sonnet is available for free on the open version of Claude AI, and businesses can customize it for their own models at a price of $3 per million input tokens and $15 per million output tokens.
Currently, the pricing for Claude 3 Haiku can only be based on tokens: $0.25 per million input tokens and $1.25 per million output tokens.
Claude 3 vs. GPT-4: First Look
Anthropic has released some performance test results for Claude 3, claiming to show comparisons between the Claude 3 series and products like ChatGPT and Google’s Gemini Ultra LLM.
Here are the performance results of Claude 3 Opus in most common AI system evaluation benchmarks, including undergraduate-level expert knowledge (MMLU), graduate-level expert reasoning (GPQA), basic mathematics (GSM8K), etc. Opus demonstrates near-human levels of understanding and fluency in complex tasks.
All Claude 3 models show enhanced capabilities in analysis and prediction, content creation, code generation, and conversing in non-English languages such as Spanish, Japanese, and French.
Below is a comparison of Claude 3 models with GPT-4 and Gemini across multiple capability benchmarks:
Of course, the key point here is that these tests were conducted by Anthropic itself and largely based on carefully designed prompts intended to leverage Claude 3’s capabilities fully.
How to Use Claude 3 Immediately
If you can’t wait to try Claude 3, the intermediate model Claude 3 Sonnet is now available on Claude AI in many countries around the world. Using Claude 3 does not require many complicated steps. Here are the specific steps:
-
Visit the Anthropic website.
-
Click the “Try Claude” button.
-
Log in with your Google account (or create a new account).
You will now enter the main Claude AI chat window, ready to start interacting with Claude 3. Opus and Sonnet are already available in the API, which is now widely accessible, allowing developers to register immediately and start using these models. Haiku will be released soon. Sonnet is supporting free experiences on claude.ai, while Opus is available for Claude Pro subscription users.
Anthropic plans to frequently update the Claude 3 model series in the coming months. It will also release a series of features to enhance the model’s capabilities, particularly for enterprise use cases and large-scale deployments. These new features will include tool usage (i.e., function calls), interactive coding (i.e., REPL), and more advanced agent capabilities.
Welcome to join the AIGC Technology Business Club Knowledge Planet
