Anthropic, the second-largest company after OpenAI, has developed a powerful family of generative AI models known as Claude. These models can perform various tasks, from captioning images and writing emails to solving math and coding challenges.
With the Anthropic model ecosystem growing so rapidly, it can be hard to keep track of what each Claude model does. To help, we’ve compiled a guide to Claude, which we’ll update as new models and upgrades are released.
Claude Models
Claude models are named after literary works: Haiku, Sonnet, and Opus. The current lineup includes:
Claude 3 Haiku – A lightweight model.
Claude 3.5 Sonnet – A mid-range model.
Claude 3 Opus – Anthropic’s flagship model.
In comparison, Claude 3.5 Sonnet, the mid-range model, is currently the most capable Claude model. The next version, Claude 3.5 Opus, is expected soon.
All Claude models can analyze images, charts, graphs, and technical diagrams from documents. They have a 200,000-token context window and can follow multi-step instructions, use tools (e.g., stock ticker trackers), and generate structured output in formats like JSON.
The "context window" refers to the amount of data a model like Claude can analyze before generating new data, while tokens are subdivisions of raw data (like breaking the word "fantastic" into smaller parts). Two hundred thousand tokens are roughly equivalent to 150,000 words or a 600-page novel.
Unlike many large generative AI models, Anthropic models don’t have internet access, meaning they’re not particularly good at answering questions about current events. They also cannot generate images—only simple line diagrams.
Regarding differences among the Claude models, Claude 3.5 Sonnet is faster and better at understanding detailed and complex instructions than Claude 3 Opus. Meanwhile, Haiku struggles with advanced prompts but is the fastest of the three.
Claude Model Pricing
Claude models are available through Anthropic’s API and managed platforms like Amazon Bedrock and Google Cloud’s Vertex AI.
Anthropic API pricing:
- Claude 3 Haiku costs $0.25 per million input tokens (~750,000 words) or $1.25 per million output tokens.
- Claude 3.5 Sonnet costs $3 per million input tokens or $15 per million output tokens.
- Claude 3 Opus costs $15 per million input tokens or $75 per million output tokens.
Anthropic offers runtime savings through instant caching and batching. Instant caching allows developers to store specific "context" for reuse in API calls, while batching processes low-priority model estimation requests in asynchronous groups (making them cheaper).
Claude’s Plans and Apps
For individual users and companies looking to interact with Claude models via apps for web, Android, and iOS, Anthropic offers a free Claude plan with rate limits and usage restrictions.
Upgrading to one of the company’s subscriptions removes these limits and unlocks new features. Current plans include:
- Claude Pro.
- Claude Team (for small businesses).
- Claude Enterprise (for large companies).
Claude Pro, priced at $20 per month, offers five times the rate limit, priority access, and previews of upcoming features.
Claude Team ($30 per user per month) adds a dashboard for billing and user management, and integrates with codebases and customer relationship management platforms (e.g., Salesforce). A toggle enables or disables references for AI-generated claims, as all models sometimes hallucinate (generate false information).
Both Pro and Team users have access to Projects, a feature that allows Claude’s output to be grounded in a knowledge base, which can include style guides, interview transcripts, and more. These users, along with free-tier users, can also use Artifacts, a workspace for editing and adding content generated by Claude, such as code, apps, or website designs.
For users who need even more, Claude Enterprise allows companies to upload proprietary data to analyze and generate insights. It also includes a larger context window (500,000 tokens), integration with GitHub for engineering teams to sync their repositories with Claude, and compatibility with Projects and Artifacts.
A Word of Caution
As with all generative AI models, there are risks associated with using Claude.
Models sometimes make mistakes when summarizing or answering questions due to their tendency to hallucinate (produce incorrect information). They are also trained on public web data, some of which may be under copyright or restricted licenses. Anthropic, like many other AI vendors, argues that the principle of fair use protects them from copyright claims, but this hasn’t stopped data owners from filing lawsuits.
Anthropic offers policies to protect specific clients from court battles stemming from fair use challenges. However, these policies don’t address the ethical concerns of using models trained on data without permission.
Posts
0 Comments