I tried out a bunch of the AI assistants. Here’s what you need to know about each one

Does it feel to you like there are way too many AI assistants to keep track of?

Between ChatGPT, Microsoft Copilot, Google Gemini, Anthropic Claude, DeepSeek, and others, it’s hard to remember what each one excels at—if anything. Beyond just the underlying differences in large language models, each AI assistant has its own features, integrations, premium features, and peculiarities.

I’m writing this guide both for myself and for anyone who wants to stay informed about generative AI. While I have some reservations, I also think it’s worth keeping an eye on what’s available. Rather than getting into the technical details of how these AI assistants work, I’ll focus on what they can actually do.

ChatGPT by OpenAI

The most recognizable name in generative AI is also the most fleshed out as a standalone consumer tech product, with features like web search, a built-in document editor, and a conversational voice mode. ChatGPT also provides one of the most robust free tiers of any AI assistant, with all but its most bleeding-edge features available in some capacity without a subscription.

Notable features:

• Voice mode for engaging in back-and-forth conversation, including an advanced version with more emotive responses.
• Traits that define how ChatGPT should converse with you.
• Temporary chats that don’t appear in your history and don’t train OpenAI’s models.
• Canvas document editor that uses AI to generate and revise content.
• Web-search for bringing in real-time information.
• GPTs that bring in information from third-party apps and services.
• Reason button that offers more in-depth problem solving and decision making.

Price: Free for limited model access, $20 per month for more advanced models and new features, $200 per month for unlimited access to experimental features.

Claude by Anthropic

As The New York Times has noted, Claude is a favorite among AI insiders for its sense of emotional intelligence, with responses that are “less like the generic prose generated by other chatbots.”

While its features are on the bare-bones side—it can’t search the web and lacks a voice conversation mode (for now)—it does offer some helpful features for creating and interacting with documents.

Notable features:

• Projects mode, which lets you upload documents and data for context in your chats.
• Artifacts, which are standalone documents and image files you can download for use in other apps.
• Styles, which let you refine how Claude writes its responses. You can even upload a document for Claude to try and mimic.

Price: Free for limited usage. $20 per month for additional models, reasoning, and the Projects feature.

Google Gemini

Google’s AI products are more diffuse than some of its startup rivals. While Gemini exists in standalone form on the web and in mobile apps—and serves as the default voice assistant on newer Android phones—it also generates summaries in Google Search, and is baked into other Google products such as Gmail, Docs, and Chrome. All this makes Gemini a bit tricky to quantify as a whole, except that it feels unavoidable if you regularly use Google’s products.

Notable features:

• Extensions for interacting with other (mostly Google) services, for instance by summarizing YouTube videos or flagging important message from Gmail.
• Gemini Live for free-flowing voice conversations.
• Google Assistant features such as smart home control and reminders.
• NotebookLM, a separate but popular product that can analyze your documents, create summaries, and even turn them into podcasts.

Price: Free, with $20 per month Gemini Advanced subscription for Workspace integration, book-length document analysis, and more advanced models.

Microsoft Copilot

Just as Gemini is built into Google products, Copilot is weaved into Microsoft’s Windows operating system, Office suite, and Edge browser. Its actual capabilities aren’t much different from other AI assistants—it primarily uses OpenAI’s large language models—but it’s easier to access if you’re deep into Microsoft’s ecosystem. (Microsoft’s GitHub also has its own version of Copilot for programmers.)

Notable features:

• Office tie-ins, including writing assistance in Word and spreadsheet analysis in Excel.
• Think Deeper offers access to OpenAI’s reasoning models.
• Copilot Voice offers free-flowing voice conversations both on desktop and mobile devices.
• Edge sidebar lets you summarize and ask about web pages.

Pricing: Free for limited usage, requires Microsoft 365 (starting at $10 per month) for full Office integration, $20 per month Copilot Pro for advanced models, early features, and Copilot in Office web apps.

DeepSeek

DeepSeek, the product of a previously-obscure Chinese company, shook up the AI world earlier this year, offering performance on par with OpenAI’s latest models with training costs that were apparently miniscule (though later disputed). It’s also raised privacy concerns over the data it sends to China, and it won’t talk about topics that are censored in China, such as the Tiananmen Square massacre.

Even so, American companies are now looking to use DeepSeek’s open-source code, including Microsoft, which is offering local DeepSeek models on Qualcomm-powered PCs, and Nvidia, which offers its own online version. As for DeepSeek’s own app, it’s a bit on the crude side, but it offers image recognition, document scanning, web search, and a “DeepThink” reasoning model for problem solving and decision making.

Notable features:

• Unlimited access to DeepSeek’s latest models.

Pricing: Free.

Grok

While Elon Musk’s AI assistant looks like a lot of its rivals on the surface, it has a kind of edgelord sensibility lurking underneath. Grok won’t flat-out tell you how to build a pipe bomb, for instance, but it won’t shut the conversation down, either. Instead, it’ll encourage you to ask more about pipe bombs in general and give you details about how they work when prompted. (Meanwhile, researchers have coaxed into providing dangerous weapon-building instructions in detail.)

Notable features:

• Voice modes with distinct personalities, including “unhinged” and “sexy” versions.
• “Think” and “DeepSearch” modes that reason through answers and pull information from online sources, respectively.

Price: Free with limited use of the latest models, $30 per month for increased rate limits and access to Think, DeepSearch, and voice modes.

Perplexity

This GPT-based AI tool started off as an alternative to traditional web search, but has since evolved into a more all-encompassing personal assistant with tools for managing documents and interacting with apps. It’s also developing its own web browser. (In other words, it’s becoming less reliant on scaping websites against their wishes.)

Notable features:

• Web search results, with summaries and citations.
• Spaces for summarizing and analyzing documents.
• Agent features on Android devices, including music playback, reminders, and calendar interactions.
• Discover section with AI-generated news summaries.

Price: Free for basic web search with some usage limits on other features, $20 per month for more deep research usage, unlimited document uploads, and a choice of AI models.

Duck.ai

With Duck.ai, DuckDuckGo offers a more privacy-focused alternative to the major AI assistants. DuckDuckGo says it’s made agreements with major AI providers so they won’t train their models on your data, and will only store it for 30 days at most. There’s no document interaction or voice chat, but it’s good enough for basic conversations.

• Private chat history that’s stored on your device, not online.
• Choice of large language models, including GPT-4o, Llama 3.3, Claude 3, o3-mini, and Mistral.
• AI answers in search results, with an option to customize how often they appear.

Price: Free

A few others

• Siri: While the current version of Siri isn’t based on large language models—and may not be for years—it will occasionally ask ChatGPT for answers on devices that support Apple Intelligence.
• Alexa+: Amazon’s just-announced AI overhaul promises more conversational capabilities than the old Alexa while still offering features like home automation, music playback, and TV suggestions. It’s launching on select Echo Show devices in an “early access period” next month.
• Meta AI: This one’s powered by Meta’s llama open-source models but is pretty basic as a consumer-facing product. It’s only available on the web or as a feature of the Meta Ray-Ban smart glasses. Standalone mobile apps are reportedly coming.

View the full article