Claude AI is a synthetic intelligence mannequin that may act as a chatbot and an AI assistant, very similar to ChatGPT and Google’s Gemini.
Named after Claude E. Shannon, generally known as the “father of knowledge concept,” Claude was designed to help with writing, coding, buyer assist and knowledge retrieval.
Claude was developed by Anthropic, a San Francisco-based firm based in 2021 by former OpenAI staff specializing in AI security and analysis. Dario Amodei, the co-founder and CEO of Anthropic, had served as vice chairman of analysis at OpenAI. His sister, Daniela Amodei, serves as Anthropic’s president.
Anthropic has drawn important funding from distinguished tech gamers. Since 2023, Amazon has invested $8 billion within the firm. As a part of the settlement, Anthropic has dedicated to utilizing Amazon Web Services as its main cloud supplier and making its AI fashions accessible to AWS prospects. The 2024 deal contains plans to increase using Amazon’s AI chips for coaching and working Anthropic’s massive language fashions. Google initially invested $500 million and plans to speculate one other $1.5 billion sooner or later.
So what can Claude do now? Here’s every part you’ll want to know, together with the fashions, plans, pricing and newest updates.
How Claude AI works
Claude AI is a flexible instrument able to answering questions, producing inventive content material like tales and poems, translating languages, transcribing and analyzing pictures, writing code, summarizing textual content and fascinating folks in pure, interactive conversations. It is out there on desktop by way of internet browsers and iOS and Android apps.
Claude makes use of massive language fashions skilled on an enormous dataset of textual content and code to know and generate human-like language.
Until not too long ago, and in contrast to different chatbots resembling OpenAI’s ChatGPT, Gemini, Copilot and Perplexity, certainly one of Claude’s largest flaws was that it could not entry real-time web or retrieve info from internet hyperlinks. Instead, generated responses based mostly solely on the information it was skilled on.
Each Claude mannequin has a particular data closing date. For instance, Claude 4 Opus and 4 Sonet had been skilled on knowledge up till March 2025 and Claude 3.5 Haiku till July 2024. Anthropic regularly up to date Claude’s knowledge to reinforce its capabilities.
Starting in March 2025, Claude lastly acquired entry to the web by a “Web Search” function. Initially, it was obtainable as a paid preview for folks within the US, after which it expanded globally in May 2025 to all Claude plans and is now obtainable on all fashions.
Knowledge cutoff nonetheless issues as a result of the net search does not substitute the coaching knowledge, however quite dietary supplements it for extra up-to-date and related responses.
Claude AI: Key options
Conversational adaptability is certainly one of its coolest options. Claude AI adjusts its tone and depth based mostly on person queries. Its skill to ask clarifying questions and keep context over prolonged exchanges makes it helpful for each informal and complicated conversations. That is among the the reason why our editors named it CNET’s greatest chatbot of 2025.
The platform additionally provides APIs that you may combine into varied instruments and workflows.
In November 2024, Anthropic launched the Model Context Protocol to its Claude desktop app, enabling the chatbot to browse the web and handle recordsdata in your laptop. This open-source protocol permits Claude to work together with varied platforms and streamline integration by eliminating the necessity for customized code.
Anthropic’s new Integrations function allows Claude to attach seamlessly with exterior apps by way of the Model Context Protocol, giving it wealthy context throughout your instruments. It can automate duties in Jira and Zapier or summarize Confluence pages — all by conversational prompts.
Another function, Advanced Research, boosts Claude’s analysis by diving deep into each inside and exterior knowledge. It can spend anyplace from 5 to 45 minutes per question and produce thorough, citation-backed reviews. It now faucets into internet searches, your individual paperwork and built-in apps like Google Workspace, supplying you with sooner, clear solutions that beforehand might have taken hours to collect manually.
Claude fashions defined
The preliminary model of Claude was launched in March 2023, adopted by Claude 2 in July 2023, permitting for extra in depth enter processing. Then, in March 2024, Anthropic launched Claude 3, comprising three fashions: Haiku, Sonnet, and Opus, every optimized for various efficiency wants.
Haiku is supposed for fast and easy duties the place velocity issues most. Sonnet balances velocity and energy for on a regular basis use, and Opus handles superior duties like arithmetic, coding and logical reasoning. In October 2024, Haiku and Sonnet had been upgraded to three.5 fashions. By May 2025, each Opus and Sonnet fashions had been upgraded to model 4.
With the discharge of the Claude 3.5 Haiku mannequin, Anthropic claims it “matches the efficiency of Claude 3 Opus” and remains to be the quickest mannequin. Anthropic additionally acknowledged within the weblog that Opus 4 is “the world’s greatest coding mannequin,” and its most clever one. Sonnet 4 delivers balanced efficiency, enhanced reasoning and extra correct responses to your directions..
In October 2024, Anthropic’s improved model of Claude 3.5 launched a beta function known as laptop use. Claude might carry out duties resembling transferring the cursor, clicking buttons and typing textual content, successfully mimicking human-computer interactions.
Claude at present helps PDF, DOCX, CSV, TXT, HTML, ODT, RTF, EPUB, JSON and XLSX. However, it has file limits inside chat uploads, resembling 30MB per file, as much as 20 recordsdata per chat and visible evaluation for PDFs below 100 pages. For detailed limits, verify Anthropic’s assist web page.
Claude at present helps PDF, DOCX, CSV, TXT, HTML, ODT, RTF, EPUB, JSON and XLSX. However, it has file limits inside chat uploads, resembling 30MB per file, as much as 20 recordsdata per chat and visible evaluation for PDFs below 100 pages. For detailed limits, verify Anthropic’s assist web page.
Claude’s ‘constitutional AI’ strategy
Claude’s distinguishing function in comparison with different generative AI fashions is its give attention to “moral” alignment and secure interactions.
On Nov. 11, 2024, Dario Amodei joined Lex Fridman for a 2-and-a-half-hour podcast to debate AI. During the dialog, he mentioned, “It is extremely unproductive to attempt to argue with another person’s imaginative and prescient.” So, he based his personal firm to display that accountable AI implementation could be each moral and worthwhile.
The “constitutional AI” framework aligns Claude’s conduct with human values. This strategy makes use of a predefined set of ideas, or “structure,” to information the AI’s responses, decreasing the chance of dangerous or biased outputs whereas guaranteeing its responses stay helpful and coherent. The structure contains tips from paperwork just like the UN’s Universal Declaration of Human Rights.
Afraz Jaffri, a senior director analyst at Gartner, advised CNET the transparency round Anthropic’s strategy “does present a sure diploma of confidence in utilization of the mannequin in environments the place responses want to achieve a excessive threshold of security, resembling in instructional settings.”
However, Jaffri cautioned that Claude customers should not rely solely on Anthropic’s built-in safeguards, and advisable utilizing exterior guardrail instruments — like these provided by different AI suppliers — to watch prompts and responses as an added layer of safety. “As not too long ago proven by their very own analysis, even with moral alignment, their Opus 4 mannequin exhibited uncharacteristic behaviour by blackmailing an engineer who had threatened to show the mannequin off,” Jaffri mentioned.
He added that restricted transparency round Claude’s coaching course of means additional security checks are nonetheless crucial.
“Every doable state of affairs can’t be absolutely coated in testing,” he advised CNET, noting that programs like Computer Use and Claude Code want extra guardrails in place to offset surprising conduct.
Claude pricing and plans
Anthropic provides quite a lot of pricing plans. There is a free possibility if you wish to check Claude with out dedication. For these looking for enhanced capabilities for particular person customers, the Claude Pro subscription is out there at $20 per 30 days. It offers enhanced utilization limits and precedence entry to new options. The Max plan provides every part in Pro for $100 per 30 days, plus extra options like extra utilization, increased limits and precedence entry throughout peak hours.
For groups, it provides an annual plan priced at $25 per member per 30 days, billed yearly, with a minimal requirement of 5 members. There can be an Enterprise plan for large-scale deployments, with custom-made pricing and options tailor-made to the precise wants of companies and organizations.
Additionally, Anthropic offers entry to its AI fashions by an API, with pricing based mostly on utilization. For instance, the Claude 3.5 Haiku mannequin is priced at 80 cents per million enter tokens and $4 per million output tokens. Tokens are textual content fragments (phrases, elements of phrases, or punctuation) that AI fashions use to course of and generate language, with pricing reflecting the quantity of knowledge dealt with.
