\n<\/p>\n
Anthropic, one of the world\u2019s largest AI vendors, has a powerful family of generative AI models called Claude. These models can perform a range of tasks, from captioning images and writing emails to solving math and coding challenges.<\/p>\n
With Anthropic\u2019s model ecosystem growing so quickly, it can be tough to keep track of which Claude models do what. To help, we\u2019ve put together a guide to Claude, which we\u2019ll keep updated as new models and upgrades arrive.<\/p>\n
Claude models are named after literary works of art: Haiku, Sonnet, and Opus. The latest are: <\/p>\n
Counterintuitively, Claude 3 Opus \u2014 the largest and most expensive model Anthropic offers \u2014 is the least capable Claude model at the moment. However, that\u2019s sure to change when Anthropic releases an updated version of Opus.<\/p>\n
Most recently, Anthropic released Claude 3.7 Sonnet<\/a>, its most advanced model to date. This AI model is different from Claude 3.5 Haiku and Claude 3 Opus because it\u2019s a hybrid AI reasoning model, which can give both real-time answers and more considered, \u201cthought-out\u201d answers to questions.<\/p>\n When using Claude 3.7 Sonnet, users can choose whether to turn on the AI model\u2019s reasoning abilities, which prompt the model to \u201cthink\u201d for a short or long period of time. <\/p>\n When reasoning is turned on, Claude 3.7 Sonnet will spend anywhere from a few seconds to a couple minutes in a \u201cthinking\u201d phase before answering. During this phase, the AI model is breaking down the user\u2019s prompt into smaller parts and checking its answers. <\/p>\n Claude 3.7 Sonnet is Anthropic\u2019s first AI model that can \u201creason,\u201d a technique\u00a0many AI labs have turned to as traditional methods of improving AI performance taper off<\/a>.<\/p>\n Even with its reasoning disabled, Claude 3.7 Sonnet remains one of the tech industry\u2019s top-performing AI models.<\/p>\n In November, Anthropic released an improved \u2013 and more expensive \u2013 version of its lightweight AI model, Claude 3.5 Haiku<\/a>. This model outperforms Anthropic\u2019s Claude 3 Opus on several benchmarks, but it can\u2019t analyze images like Claude 3 Opus or Claude 3.7 Sonnet can.<\/p>\n All Claude models \u2014 which have a standard 200,000-token context window \u2014 can also follow multistep instructions, use tools<\/a> (e.g., stock ticker trackers), and produce structured output in formats like\u00a0JSON<\/a>.<\/p>\n A context window is the amount of data a model like Claude can analyze before generating new data, while tokens are subdivided bits of raw data (like the syllables \u201cfan,\u201d \u201ctas,\u201d and \u201ctic\u201d in the word \u201cfantastic\u201d). Two hundred thousand tokens is equivalent to about 150,000 words, or a 600-page novel.<\/p>\n Unlike many major generative AI models, Anthropic\u2019s can\u2019t access the internet, meaning they\u2019re not particularly great at answering current events questions. They also can\u2019t generate images \u2014 only simple line diagrams. <\/p>\n As for the major differences between Claude models, Claude 3.7 Sonnet is faster than Claude 3 Opus and better understands nuanced and complex instructions. Haiku struggles with sophisticated prompts, but it\u2019s the swiftest of the three models.<\/p>\n The Claude models are available through Anthropic\u2019s API and managed platforms such as Amazon Bedrock<\/a> and Google Cloud\u2019s Vertex AI<\/a>. <\/p>\n Here\u2019s the Anthropic API pricing:<\/p>\n Anthropic offers prompt caching and batching to yield additional runtime savings. <\/p>\n Prompt caching lets developers store specific \u201cprompt contexts\u201d that can be reused across API calls to a model, while batching processes asynchronous groups of low-priority (and subsequently cheaper) model inference requests.<\/p>\n For individual users and companies looking to simply interact with the Claude models via apps for the web, Android, and iOS, Anthropic offers a free Claude plan with rate limits and other usage restrictions. <\/p>\n Upgrading to one of the company\u2019s subscriptions removes those limits and unlocks new functionality. The current plans are:<\/p>\n Claude Pro, which costs $20 per month, comes with 5x higher rate limits, priority access, and previews of upcoming features.<\/p>\n Being business-focused, Team \u2014 which costs $30 per user per month \u2014 adds a dashboard to control billing and user management and integrations with data repos such as codebases and customer relationship management platforms (e.g., Salesforce). A toggle enables or disables citations to verify AI-generated claims. (Like all models, Claude hallucinates<\/a> from time to time.)<\/p>\n Both Pro and Team subscribers get Projects, a feature that grounds Claude\u2019s outputs in knowledge bases, which can be style guides, interview transcripts, and so on. These customers, along with free-tier users, can also tap into Artifacts, a workspace where users can edit and add to content like code, apps, website designs, and other docs generated by Claude.<\/p>\n For customers who need even more, there\u2019s Claude Enterprise, which allows companies to upload proprietary data into Claude so that Claude can analyze the info and answer questions about it. Claude Enterprise also comes with a larger context window (500,000 tokens), GitHub integration for engineering teams to sync their GitHub repositories with Claude, and Projects and Artifacts.<\/p>\n As is the case with all generative AI models, there are risks associated with using Claude.<\/p>\n The models occasionally\u00a0make mistakes when summarizing<\/a>\u00a0or answering questions<\/a> because of their\u00a0tendency to\u00a0hallucinate<\/a>. They\u2019re also trained on public web data, some of which may be copyrighted or under a restrictive license. Anthropic and many other AI vendors argue that the\u00a0fair-use<\/a>\u00a0doctrine shields them from copyright claims. But that hasn\u2019t stopped data owners from<\/a>\u00a0filing lawsuits<\/a>.<\/p>\nClaude model pricing<\/h2>\n
\n
Claude plans and apps<\/h2>\n
A word of caution<\/h2>\n