One of the world’s largest AI vendors, humanity has a strong family of generative AI models called Claude. These models can perform a variety of tasks, from captioning images and writing emails to solving mathematics and coding challenges.
With the Anthropic model ecosystem growing so rapidly it can be difficult to track which Claude models do what. To help, we have compiled a guide to Claude. This will continue to be updated as new models and upgrades arrive.
Claude model
The Claude model is named after the works of art from literary works: haiku, sonnets and opas. The latest ones are:
Claude 3.5 Haiku, lightweight model. Claude 3.7 Sonnet, midrange, hybrid reasoning model. This is currently the flagship AI model of humanity. Claude 3 Opus, a big model.
Counterintuitively, the Claude 3 Opus (the biggest and most expensive model human offer) is the lowest-capacity Claude model at the moment. However, this will definitely change when humanity releases an updated version of Opus.
Recently, humanity has released its Claude 3.7 Sonnet, the most advanced model ever. Unlike the Claude 3.5 Haiku and Claude 3 Opus, this AI model is a hybrid AI inference model that can provide both real-time answers and “conceived” answers to questions.
When using Claude 3.7 Sonnet, users can choose whether to turn on the inference capabilities of the AI model. This will make you “think” in the model for a short or long period of time.
When inference is turned on, Claude 3.7 Sonnet spends seconds or minutes in the “thinking” stage before answering. In this phase, the AI model breaks down the user’s prompt into smaller pieces and checks the answer.
Claude 3.7 Sonnet is the first AI model for humanity that can “infer.” This is a technique that many AI labs have seen as a traditional way to improve AI performance taper.
Even if that inference is disabled, Claude 3.7 Sonnet continues to be one of the top-performing AI models in the tech industry.
In November, humanity released an improved, more expensive version of the lightweight AI model, the Claude 3.5 Haiku. This model outperforms Anthropic’s Claude 3 Opus in some benchmarks, but it cannot analyze images such as Claude 3 Opus or Claude 3.7 Sonnet Can.
All Claude models with standard 200,000 token context windows can follow multi-step instructions and also use tools (such as stock ticker trackers) to generate structured output in the form of JSON or similar.
A context window is the amount of data that a model like Claude can analyze before generating new data, and tokens are “subdivided bits of raw data (“fan”, “TAS”, “TIC” etc. Like “fantastic”.) Two million tokens are equivalent to a novel with around 150,000 words or 600 pages.
Unlike many major generator AI models, Anthropic does not have internet access. That is, it’s not particularly good for questions about current events. Furthermore, it is not possible to generate images. It’s just a simple diagram.
As for the big differences between Claude models, Claude 3.7 Sonnet is faster than the Claude 3 Opus and better understands subtle and complex instructions. Haiku struggles with sophisticated prompts, but it is the fastest of the three models.
Claude model pricing
The Claude model is available through Anthropic’s API and managed platforms such as Amazon Bedrock and Google Cloud’s Vertex AI.
This is human API pricing:
Claude 3.5 Haiku costs 1 million input tokens (~750,000 words) or 4 output tokens per million dollars 3.7 cost 3.7 $3 per sonnet input token, or 15 output tokens per dollar Claud 100 1,000 output tokens
Humanity offers rapid caching and batching, bringing additional runtime savings.
Prompt caching allows developers to store specific “prompt contexts” that can be reused throughout the API call to the model, but batching processes asynchronous groups of low-priority (and then cheap) model inference requests .
Claude Plan and the App
For individual users and businesses who are simply trying to interact with the Claude model via apps for the web, Android and iOS, Anthropic offers a free Claude plan with fee limits and other usage restrictions.
Upgrading to one of your company’s subscriptions removes these restrictions and unlocks new features. The current plans are as follows:
At $20 a month, Claude Pro offers 5x rate limits, priority access and previews of upcoming features.
Business-focused teams (which cost $30 per user per month) are dashed to control billing, user management and integration with data repositories such as codebases and customer relationship management platforms (such as Salesforce). Add the board. The toggle enables or disables citations and reviews claims generated by AI. (Leod, like all models, sometimes hallucinates.)
Both PRO and team subscribers will get the project. This is a feature based on Claude’s knowledge base output. This can be a style guide, interview transcript, and more. These customers can also leverage Artifacts, a workspace that users can edit and add to content such as code, apps, website designs, and other Claude-generated documents.
Even more customers have Claude Enterprise. This allows businesses to upload their own data to Claude, allowing Claude to analyze the information and answer questions. Claude Enterprise comes with a larger context window (500,000 tokens), GitHub integration for the Engineering team to synchronize Claude and GitHub repository, and projects and artifacts.
Words of caution
As with all generation AI models, there are risks associated with using Claude.
Models sometimes make mistakes when summarizing or answering questions due to hallucination tendencies. We are also trained on public web data, some of which may be copyrighted or under a limited license. Humanity and many other AI vendors argue that fair doctrine protects them from copyright claims. But that doesn’t stop data owners from filing lawsuits.
Humanity provides policies that protect certain customers from the court battles arising from fair challenges. However, it does not solve the ethical difficulties of using data-trained models without permission.
This article was originally published on October 19, 2024. Updated on February 25th, 2025, and includes new details for Claude 3.7 Sonnet and Claude 3.5 Haiku.