§6

Consumer AI assistants and AI search

This category covers the general-purpose chat assistants and answer engines that ordinary people and knowledge workers use directly, as opposed to APIs or coding agents. As of September 2026 the market has consolidated around five subscription ladders (OpenAI, Anthropic, Google, Microsoft, xAI) priced from roughly $5 to $300 per month, plus a second tier of answer engines (Perplexity), agentic "do the work" products (Manus, Genspark, Kimi), and cheap or free open-weight-backed challengers (DeepSeek, Mistral).

Data checked 2026-09-06

Two structural shifts define the current moment: pricing has moved from flat message caps to compute- or credit-metered usage, and the assistant is migrating out of the chat box into browsers (Comet, ChatGPT Atlas) and into OS/office suites. Naming churn is severe — NotebookLM is now Gemini Notebook, Le Chat is now Mistral Vibe, Copilot Pro has been folded into Microsoft 365 Premium, and You.com has abandoned the consumer assistant entirely for search APIs.

A How to choose

Start with where your data already lives, because integration beats raw model quality for daily use: if your documents are in Microsoft 365, Copilot is effectively free with a subscription you already pay for; if they are in Gmail/Drive, Gemini's connectors do work no third party can replicate; only if you are platform-agnostic does the model itself become the deciding axis. The second axis is what you are actually buying — a fast answer with citations (Perplexity, Duck.ai), a long reasoning session (Claude, ChatGPT Pro, Gemini Ultra), or an autonomous agent that produces finished files while you close the laptop (Manus, Genspark, Kimi); these are genuinely different products and paying frontier-assistant prices for citation-backed search is waste, while asking a search engine to do multi-hour agentic work will fail. The third axis is cost structure: flat-rate plans like ChatGPT Plus ($19.99/mo) and Claude Pro ($20/mo) are predictable, whereas credit-metered products (Manus, Genspark, Perplexity credit packs) can quietly cost several hundred dollars a month once agents start running, so budget by task volume rather than seat count.

The fourth is data governance — DeepSeek and Kimi are the cheapest capable options by a wide margin but process data in China, which is disqualifying for most regulated work; Proton Lumo and Duck.ai are the credible choices when the content itself must not be retained. Do not buy the popular default reflexively: the $100–$300 top tiers (ChatGPT Pro, Claude Max, Grok Heavy, Genspark Pro) are only rational if you are running long agentic sessions daily and can point to hours saved, and most individuals are better served by a $20 plan plus occasional credit top-ups. Also resist paying for two overlapping frontier assistants; the more common gap is between one good assistant and one specialised tool — a research notebook (Gemini Notebook) or a model-router (Poe) — rather than a second chatbot.

Finally, treat browser-resident assistants as a separate commitment: Comet and similar products want your logged-in sessions and full browsing context, which is a much larger trust decision than pasting text into a chat window.

B At a glance

Name	Price tiers	Web access	File upload	Voice	Platforms	Pricing
ChatGPT	Free / Go $8 / Plus $19.99 / Pro 5x $100 / Pro 20x $200 per month	Yes — built-in browsing plus deep research; also available in the ChatGPT Atlas desktop browser	Yes — PDFs, images, spreadsheets, code; free tier limited	Yes — Advanced Voice Mode, real-time, with camera sharing on mobile	Web, iOS, Android, macOS, Windows, Atlas browser	Free tier; ChatGPT Go $8.00/mo; Plus $19.99/mo (~$200/yr); Pro 5x $100.00/mo; Pro 20x $200.00/mo; credit packs $4.00/100, $20.00/500, $40.00/1000. Prices read from the official OpenAI iOS listing (App Store IAP range $8.00–$200.00 corroborated on Google Play); openai.com/chatgpt/pricing blocks automated fetches, so tier naming is from store metadata rather than the pricing page.
Claude	Free / Pro $20 (or $17 annual) / Max 5x $100 / Team $25 seat / Enterprise per month	Yes — web search on all tiers including free, plus research mode	Yes — PDFs, images, spreadsheets, code; file creation with code execution	Yes — voice mode on mobile apps; no real-time video/camera equivalent	Web, iOS, Android, macOS, Windows, CLI (Claude Code)	Free $0; Pro $17/mo billed annually or $20/mo monthly; Max 5x $100/mo; Team $20/seat/mo annual ($25 monthly), premium seat $100/mo; Enterprise $20/seat/mo plus API-rate usage. Max 20x is sold above the 5x tier (iOS in-app price $249.99 including Apple's markup); its exact web price was not visible on the public pricing page.
Google Gemini	Free / AI Plus $4.99 / AI Pro $19.99 / AI Ultra $99.99–$199.99 per month	Yes — Google Search grounding plus Deep Research and Deep Search	Yes — documents, spreadsheets, photos, video	Yes — Gemini Live with real-time screen and camera sharing	Web, iOS, Android, Chrome, Google Workspace apps, Android system assistant	Free $0; Google AI Plus $4.99/mo (400GB, 2x free limits); Google AI Pro $19.99/mo (5TB, 4x free limits); Google AI Ultra from $99.99/mo (5x Pro) to $199.99/mo (20x Pro, 20–30TB). Business: Gemini Enterprise from $21/seat/mo, Standard/Plus from $30/seat/mo.
Microsoft Copilot	Free / M365 Basic $1.99 / Personal $9.99 / Family $12.99 / Premium $19.99 per month	Yes — Bing-grounded search and citations on all tiers	Yes — Word, Excel, PowerPoint, PDF and images from web and mobile	Yes — voice chat in the Copilot apps	Web, iOS, Android, Windows (built in), Microsoft 365 apps, Edge	Copilot free; Microsoft 365 Basic $1.99/mo; Personal $9.99/mo or $99.99/yr; Family $12.99/mo or $129.99/yr; Premium $19.99/mo or $199.99/yr (agentic features, highest Copilot usage). Copilot Pro as a standalone SKU no longer appears in Microsoft's consumer plan comparison or the Copilot app's in-app purchase list.
Perplexity	Free / Pro $20 / Max $200 per month, plus credit packs	Yes — this is the core product; live search with inline citations on every answer	Yes — PDFs, images, spreadsheets; higher limits on Pro	Yes — voice input and spoken answers in the mobile apps	Web, iOS, Android, macOS, Windows, Comet browser	Free tier; Pro $20/mo or $200/yr; Max $200/mo; credit packs $5/500, $10/1,000, $25/2,500, $50/5,000, $100/10,000. Prices read from the official Perplexity iOS listing; perplexity.ai/pricing blocks automated fetches.
Comet	Free with Perplexity account; capability scales with Pro $20 / Max $200 per month	Yes — it is a browser; agent reads live and authenticated pages	Yes — via the Perplexity assistant sidebar	Yes — voice via the Perplexity assistant	macOS, Windows, iOS	Bundled with a Perplexity account; free to install, with agentic capacity governed by the Perplexity tier (Free / Pro $20/mo / Max $200/mo). No separate Comet-only price was verifiable — comet.perplexity.ai blocks automated fetches.
Grok	Free / SuperGrok Lite $10 / SuperGrok $30 / SuperGrok Plus $100 / Heavy $300 per month	Yes — web search plus real-time X post search and DeepSearch	Yes — documents, photos and pasted links for analysis	Yes — voice mode with live camera sharing and selectable personalities	Web, iOS, Android, X app, Telegram	Free tier in the Grok app and on X; SuperGrok Lite $10.00/mo; SuperGrok $30.00/mo; SuperGrok Plus $100.00/mo; SuperGrok Heavy $300.00/mo; extra usage credit packs $5–$100. Prices read from the official xAI iOS listing (Play Store IAP range corroborates $8.00–$200.00 for consumables); x.ai blocks automated fetches so the web pricing page could not be confirmed.
Meta AI	Free only — no paid consumer tier	Yes — web search plus Meta's own social and reviews data	Yes — photos, docs, PDFs and spreadsheets, multiple files at once	Yes — voice chat in-app and hands-free on Ray-Ban Meta glasses	WhatsApp, Instagram, Messenger, Facebook, web, iOS, Android, Ray-Ban Meta glasses	$0 — free across WhatsApp, Instagram, Messenger, Facebook, meta.ai and the standalone app; the official Meta AI iOS listing shows no in-app purchases. No consumer subscription is offered.
DeepSeek	Free app; API metered per token	Yes — web search in the chat app	Yes — documents and images, though limits are not publicly documented	Limited — voice input on mobile; no full conversational voice mode comparable to rivals	Web, iOS, Android, API; open weights self-hostable	Consumer chat app: $0, no in-app purchases on the official iOS listing. API is paid and metered per million tokens (see api-docs.deepseek.com/quick_start/pricing); exact current per-token rates were not verified for this entry.
Mistral Vibe (formerly Le Chat)	Free / Pro $14.99 (student $5.99) / Team $24.99 per user / Enterprise per month	Yes — web search with verified news sourcing; limited on free tier	Yes — spreadsheets, databases, PDFs, plus 100+ tool connectors	Yes — voice input powered by Mistral's Voxtral model	Web, iOS, Android, CLI, VS Code, JetBrains, Zed	Free $0 (limited messages, web searches and coding sessions, $10/mo API credits); Pro $14.99/mo excl. tax ($5.99/mo verified students, max 12 months) with $15–$30/mo API credits; Team $24.99/user/mo excl. tax with a $50/mo minimum and 30GB storage per user; Enterprise on request.
Kimi	Free / Moderato $19 / Allegretto $39 / Allegro $99 / Vivace $199 per month	Yes — web research plus data-source plugins (Wind, S&P)	Yes — PPT, Word, Excel, PDF and images, with lossless format conversion	Limited — voice input on mobile; no headline conversational voice mode	Web, iOS, Android, desktop; open weights self-hostable	Free tier; Moderato $19.00/mo ($180.99/yr); Allegretto $39.00/mo ($374.99/yr); Allegro $99.00/mo ($949.00/yr); Vivace $199.00/mo ($1,999.00/yr). Prices read from the official Moonshot iOS listing; kimi.com's pricing page renders client-side and could not be scraped, and CNY-denominated domestic pricing may differ.
Manus	Free chat / Basic $19 / Plus $39 / Pro $199 per month, plus credit packs from $20	Yes — autonomous browsing and wide research with parallel sub-agents	Yes — Excel, CSV and documents; produces spreadsheets, decks and images	Limited — mobile voice input; no conversational voice mode	Web, iOS, Android, desktop app, Slack integration, browser operator	Free chat; Manus Basic $19.00/mo; Plus $39.00/mo; Pro $199.00/mo; credit top-ups from $20.00 for 2,000 credits. Prices read from the official Manus iOS listing; manus.im/pricing renders client-side and its tables could not be scraped.
Gemini Notebook (formerly NotebookLM)	Free / capacity via Google AI Plus $4.99, Pro $19.99, Ultra $99.99–$199.99 per month	Partial — can discover and import web sources, but answers are grounded only in the notebook	Yes — PDFs, websites, YouTube videos, Docs and pasted text as sources	Yes — generated Audio Overviews with interactive join-in; background and offline playback	Web, iOS, Android; notebooks also surface in Gemini apps and Google Search AI Mode	Free $0 at standard limits; capacity scales with Google AI Plus $4.99/mo (2x standard), AI Pro $19.99/mo (4x standard) and AI Ultra $99.99–$199.99/mo (5x or 20x AI Pro). From 2 September 2026 limits are compute-based, factoring prompt complexity, models and features, refreshing every five hours up to a weekly cap. Google no longer publishes fixed per-notebook or per-day counts.
Genspark	Free / Plus $24.99–$99.99 by credit tier / Pro $249.99 per month	Yes — autonomous research and information gathering by the Super Agent	Yes — documents and data for sheets, docs and deck generation; GitHub integration	Yes — meeting transcription and AI Pods audio generation; voice input in-app	Web, iOS, Android, browser extension	Free tier; Plus $24.99/mo (10K credits), $49.99/mo (21K), $74.99/mo (32K), $99.99/mo (44K); Plus annual $239.99; Pro $249.99/mo; credit packs $19.99 and $39.99. Prices read from the official Genspark iOS listing; genspark.ai/pricing blocks automated fetches.
Poe	Free / Basic $4.99 / Plus $19.99 per month (annual $49.99 / $199.99)	Partial — depends on the selected bot; not a uniform capability across models	Yes — file and image attachments, subject to per-model support	Limited — no first-class conversational voice mode comparable to the majors	Web, iOS, Android, macOS, Windows	Free tier; Basic $4.99/mo or $49.99/yr; Plus $19.99/mo or $199.99/yr. Prices read from the official Quora/Poe iOS listing; poe.com/pricing renders client-side and the point allowances per tier could not be verified.
Proton Lumo	Free / Lumo Plus paid monthly — exact USD price not published on the product page	Yes — optional web search, off by default in keeping with the privacy model	Yes — file analysis, with Proton Drive integration	Limited — no full conversational voice mode documented	Web, iOS, Android; Tor guest mode without an account	Free tier (limited Max-model access, limited messages and chat history, limited image generation, one project). Lumo Plus is a paid monthly subscription with a 30-day money-back guarantee; the exact USD price is not stated on proton.me/lumo, which shows currency-dependent pricing only — treat the number as unknown.
Duck.ai	Free — no paid Duck.ai tier confirmed	Partial — integrated with DuckDuckGo search results, not a full browsing agent	Limited — not a documented core feature; treat as unsupported for real workflows	No	Web, DuckDuckGo browser (macOS, Windows, iOS, Android)	$0 — the page presents Duck.ai as free with no paid tier shown. DuckDuckGo has previously offered a paid higher-limit option bundled with its subscription products; no current paid Duck.ai tier was verifiable, so treat paid availability as unknown.
Qwen Studio (formerly Qwen Chat)	unknown — free access with an account; paid tiers not publicly listed	Yes — web search in the chat interface	Yes — documents and images	unknown — not documented on the public product page	Web, iOS, Android (region-dependent); open weights self-hostable	unknown — chat.qwen.ai requires a login and renders its plan information client-side; no consumer subscription prices were retrievable. Free access with an account is confirmed from the product page.
You.com	Consumer plans discontinued; APIs from $5 CPM (per 1,000 search queries)	Yes — this is now the entire product, sold as an API	No — no consumer assistant remains to upload files to	No	API only (Python SDK, REST); no consumer apps	Web Search API advertised at $5 CPM (per thousand queries); the site markets it as roughly 35% cheaper than Claude's built-in web search. No consumer assistant subscription is offered any more; per-API rate cards are on you.com/pricing.

C Entries

ChatGPT

OpenAI's consumer assistant, now running the GPT-6 'Astra' family alongside the GPT-5.6 line. It bundles web browsing, code execution, image generation, voice, connectors to Gmail/Drive/SharePoint, custom GPTs, and agentic task execution into one subscription ladder. The paid ladder has widened rather than deepened: a cheap $8 Go tier for emerging markets and light users, Plus at $19.99, and Pro split into 5x ($100) and 20x ($200) usage variants, plus consumable credit packs ($4 for 100 credits) for overflow. Distinguishing feature versus Claude and Gemini is breadth — it is the only one of the three that is simultaneously competitive at image generation, voice, agents, and general chat rather than strongest at one.

Price tiers	Free / Go $8 / Plus $19.99 / Pro 5x $100 / Pro 20x $200 per month
Web access	Yes — built-in browsing plus deep research; also available in the ChatGPT Atlas desktop browser
File upload	Yes — PDFs, images, spreadsheets, code; free tier limited
Voice	Yes — Advanced Voice Mode, real-time, with camera sharing on mobile
Platforms	Web, iOS, Android, macOS, Windows, Atlas browser

Watch out: Usage limits are opaque and change without notice, and the tier structure is now confusing enough (Go / Plus / Pro 5x / Pro 20x / credits) that it is hard to predict what you will be able to do for a given price. Memory and personalisation are on by default and leak between contexts in ways that surprise people. The wrong choice when you need auditable citations for every claim — it browses, but its sourcing discipline is weaker than Perplexity's — or when your employer forbids consumer-tier data handling.

Free tier; ChatGPT Go $8.00/mo; Plus $19.99/mo (~$200/yr); Pro 5x $100.00/mo; Pro 20x $200.00/mo; credit packs $4.00/100, $20.00/500, $40.00/1000. Prices read from the official OpenAI iOS listing (App Store IAP range $8.00–$200.00 corroborated on Google Play); openai.com/chatgpt/pricing blocks automated fetches, so tier naming is from store metadata rather than the pricing page.

Claude

Anthropic's assistant, running the Opus 5 and Fable 5.1 model families. The free tier now includes web search, memory, file creation with code execution, Skills, connectors and incognito chats. Pro at $20/mo ($17 billed annually) adds Claude Code, Claude Cowork, Claude Design and Microsoft 365 integration; Max at $100/mo buys 5x Pro usage per session with a higher 20x tier above it. Compared with ChatGPT it is narrower — no native image generation, weaker voice — but stronger on long documents, code review and instruction-following over multi-turn sessions.

Price tiers	Free / Pro $20 (or $17 annual) / Max 5x $100 / Team $25 seat / Enterprise per month
Web access	Yes — web search on all tiers including free, plus research mode
File upload	Yes — PDFs, images, spreadsheets, code; file creation with code execution
Voice	Yes — voice mode on mobile apps; no real-time video/camera equivalent
Platforms	Web, iOS, Android, macOS, Windows, CLI (Claude Code)

Watch out: No native image or video generation, and voice is far behind ChatGPT and Gemini. Session-based usage caps on Pro bite hard on long coding or document sessions — hitting a five-hour window limit mid-task is a routine experience and the fix is a $100/mo upgrade. The wrong choice if you want one assistant to also make images, run voice conversations, or search your personal Google data.

Free $0; Pro $17/mo billed annually or $20/mo monthly; Max 5x $100/mo; Team $20/seat/mo annual ($25 monthly), premium seat $100/mo; Enterprise $20/seat/mo plus API-rate usage. Max 20x is sold above the 5x tier (iOS in-app price $249.99 including Apple's markup); its exact web price was not visible on the public pricing page.

Google Gemini

Google's assistant, currently on Gemini 3.8 Flash and Gemini Omni, sold through the Google AI plan ladder rather than as a standalone subscription — the same $19.99/mo AI Pro plan that gives you 5TB of Drive storage also raises your Gemini limits, bundles YouTube Premium Lite, the Jules coding agent and Gemini Notebook. The differentiator is data gravity: connectors into Gmail, Calendar, Photos, YouTube and Search do work no competitor can replicate, and the Gemini Spark agent and Daily Brief run proactively against that data. Image generation runs through Nano Banana 2 and video through Veo.

Price tiers	Free / AI Plus $4.99 / AI Pro $19.99 / AI Ultra $99.99–$199.99 per month
Web access	Yes — Google Search grounding plus Deep Research and Deep Search
File upload	Yes — documents, spreadsheets, photos, video
Voice	Yes — Gemini Live with real-time screen and camera sharing
Platforms	Web, iOS, Android, Chrome, Google Workspace apps, Android system assistant

Watch out: From September 2026 limits are compute-based — they vary with prompt complexity, model, chat length and feature, refreshing every five hours up to a weekly cap — which makes it genuinely hard to predict when you will run out. The plans are bundles, so you are paying for Drive storage and YouTube whether or not you want them. Model routing is opaque: you often do not know whether Flash or Pro answered you, and quality varies accordingly. Wrong choice if you want a clean, auditable assistant divorced from Google's consumer data graph.

Free $0; Google AI Plus $4.99/mo (400GB, 2x free limits); Google AI Pro $19.99/mo (5TB, 4x free limits); Google AI Ultra from $99.99/mo (5x Pro) to $199.99/mo (20x Pro, 20–30TB). Business: Gemini Enterprise from $21/seat/mo, Standard/Plus from $30/seat/mo.

Microsoft Copilot

Microsoft's consumer assistant is free standalone and gets progressively more capable as you climb the Microsoft 365 subscription ladder — Personal at $9.99/mo, Family at $12.99/mo, and Microsoft 365 Premium at $19.99/mo, which absorbed the old standalone Copilot Pro tier. It reads and writes Word, Excel, PowerPoint and Outlook content directly, offers model choice between several underlying LLMs, and adds Cowork, an agentic mode that executes multi-step work and returns finished files. Its value proposition is not model quality but that the assistant sits inside the documents rather than beside them.

Price tiers	Free / M365 Basic $1.99 / Personal $9.99 / Family $12.99 / Premium $19.99 per month
Web access	Yes — Bing-grounded search and citations on all tiers
File upload	Yes — Word, Excel, PowerPoint, PDF and images from web and mobile
Voice	Yes — voice chat in the Copilot apps
Platforms	Web, iOS, Android, Windows (built in), Microsoft 365 apps, Edge

Watch out: Copilot has a long record of being a slower, more conservative wrapper over models you can access more directly elsewhere, and quality varies sharply between the free chat surface, the Office panes and Cowork. Copilot in Family plans is limited to the subscription owner. The tier map is genuinely confusing after the Copilot Pro retirement, and Microsoft 365 Premium overlaps heavily with what Personal already provides. Wrong choice if you want frontier reasoning quality per dollar rather than Office integration.

Copilot free; Microsoft 365 Basic $1.99/mo; Personal $9.99/mo or $99.99/yr; Family $12.99/mo or $129.99/yr; Premium $19.99/mo or $199.99/yr (agentic features, highest Copilot usage). Copilot Pro as a standalone SKU no longer appears in Microsoft's consumer plan comparison or the Copilot app's in-app purchase list.

Perplexity

An AI search product rather than a chatbot: every answer carries numbered source links, and the interface is built around follow-up questions on a research thread rather than open-ended conversation. Pro at $20/mo unlocks Pro Search, Deep Research, Labs (which builds reports and small apps), and model choice across OpenAI, Anthropic and other vendors; Max at $200/mo raises limits and bundles the Comet browser experience. Consumable credit packs ($5 for 500 up to $100 for 10,000) meter heavy agentic use on top of the subscription.

Price tiers	Free / Pro $20 / Max $200 per month, plus credit packs
Web access	Yes — this is the core product; live search with inline citations on every answer
File upload	Yes — PDFs, images, spreadsheets; higher limits on Pro
Voice	Yes — voice input and spoken answers in the mobile apps
Platforms	Web, iOS, Android, macOS, Windows, Comet browser

Watch out: It is a worse writing and coding assistant than any of the frontier chat products, because it optimises for retrieving and citing rather than reasoning at length. Citations are present but not always load-bearing: the model still occasionally states things the cited page does not support, so the links must be clicked. The $200/mo Max tier is hard to justify against a $20 Pro plan for most users. Perplexity has also drawn repeated publisher complaints about crawling and content use, which matters if your organisation cares about that.

Free tier; Pro $20/mo or $200/yr; Max $200/mo; credit packs $5/500, $10/1,000, $25/2,500, $50/5,000, $100/10,000. Prices read from the official Perplexity iOS listing; perplexity.ai/pricing blocks automated fetches.

Comet

Perplexity's AI browser, distributed as a desktop and iOS app, which puts an assistant sidebar alongside every page and lets it read, summarise and act on what you are looking at — including sites you are authenticated to. It handles cross-tab summarisation, shopping and booking flows, and scheduled chores rather than only answering questions. Access comes with a Perplexity account, with the deepest capabilities gated behind Pro and Max, so it is best understood as a distribution surface for the Perplexity subscription rather than a separate purchase.

Price tiers	Free with Perplexity account; capability scales with Pro $20 / Max $200 per month
Web access	Yes — it is a browser; agent reads live and authenticated pages
File upload	Yes — via the Perplexity assistant sidebar
Voice	Yes — voice via the Perplexity assistant
Platforms	macOS, Windows, iOS

Watch out: Handing a hosted agent your logged-in browser sessions is a materially larger trust decision than pasting text into a chatbot, and prompt-injection from a hostile page is a real and demonstrated attack class against agentic browsers. It is also yet another Chromium browser you must migrate into, with extension and profile friction. Wrong choice for anyone whose browser holds credentials for systems they cannot afford an agent to misuse.

Bundled with a Perplexity account; free to install, with agentic capacity governed by the Perplexity tier (Free / Pro $20/mo / Max $200/mo). No separate Comet-only price was verifiable — comet.perplexity.ai blocks automated fetches.

Grok

xAI's assistant, currently on the Grok 4.x line, differentiated by real-time access to X (Twitter) posts alongside web search — useful for breaking news and sentiment, and unavailable at that latency anywhere else. It bundles Grok Imagine for fast image and short video generation with sound, a voice mode with live camera, and configurable personalities including character companions. The subscription ladder is unusually wide: SuperGrok Lite at $10/mo through SuperGrok Heavy at $300/mo, and Grok is also included with X Premium+ subscriptions.

Price tiers	Free / SuperGrok Lite $10 / SuperGrok $30 / SuperGrok Plus $100 / Heavy $300 per month
Web access	Yes — web search plus real-time X post search and DeepSearch
File upload	Yes — documents, photos and pasted links for analysis
Voice	Yes — voice mode with live camera sharing and selectable personalities
Platforms	Web, iOS, Android, X app, Telegram

Watch out: Content moderation is deliberately looser, which produces both the appeal and the liability — Grok has repeatedly generated output that would be a serious incident in a corporate context, and its companion characters are not something most workplaces want on managed devices. Enterprise governance, compliance attestations and admin controls lag the other majors. Its live X grounding also inherits X's information quality, which is a poor primary source. Wrong choice for regulated or brand-sensitive work.

Free tier in the Grok app and on X; SuperGrok Lite $10.00/mo; SuperGrok $30.00/mo; SuperGrok Plus $100.00/mo; SuperGrok Heavy $300.00/mo; extra usage credit packs $5–$100. Prices read from the official xAI iOS listing (Play Store IAP range corroborates $8.00–$200.00 for consumables); x.ai blocks automated fetches so the web pricing page could not be confirmed.

Meta AI

Meta's assistant, distributed through the apps billions of people already have plus a standalone app and Ray-Ban smart glasses. It handles document, PDF and spreadsheet analysis, image generation and editing, structured research reports, and now agentic tasks like daily briefings, reminders, shopping lists and email/calendar handling. Its distinguishing input is Meta's own social graph — it draws on posts, reviews and creators across Instagram, Facebook and Threads for recommendations and local suggestions, which no other assistant can do. There is no paid consumer tier; the standalone iOS app lists no in-app purchases.

Price tiers	Free only — no paid consumer tier
Web access	Yes — web search plus Meta's own social and reviews data
File upload	Yes — photos, docs, PDFs and spreadsheets, multiple files at once
Voice	Yes — voice chat in-app and hands-free on Ray-Ban Meta glasses
Platforms	WhatsApp, Instagram, Messenger, Facebook, web, iOS, Android, Ray-Ban Meta glasses

Watch out: Materially weaker than ChatGPT, Claude or Gemini on reasoning, code and long documents, and it is free because your interactions feed Meta's advertising and recommendation systems — Meta has stated AI chat signals inform ad targeting. Availability of specific features varies by country and regulatory regime, and it is absent or degraded in parts of the EU. There is no paid tier to escape usage limits or data use. Wrong choice for any work you would not post publicly.

$0 — free across WhatsApp, Instagram, Messenger, Facebook, meta.ai and the standalone app; the official Meta AI iOS listing shows no in-app purchases. No consumer subscription is offered.

DeepSeek

DeepSeek's consumer app gives free access to its current flagship, DeepSeek-V4-Pro, which the company positions around agent capabilities, Responses API support and Codex integration. The consumer product is deliberately spartan — chat, web and mobile — with the real strategic value in the accompanying open-weight model releases and an API priced far below Western frontier vendors. For practitioners it is the reference point for how much capability is available at near-zero cost, and the models can be self-hosted, which none of the US frontier assistants permit.

Price tiers	Free app; API metered per token
Web access	Yes — web search in the chat app
File upload	Yes — documents and images, though limits are not publicly documented
Voice	Limited — voice input on mobile; no full conversational voice mode comparable to rivals
Platforms	Web, iOS, Android, API; open weights self-hostable

Watch out: Data is processed in China under Chinese law, which disqualifies it for most regulated, government and enterprise work regardless of quality, and several governments have banned it on official devices. The consumer app is feature-poor next to ChatGPT or Gemini — thin voice, thin connectors, no meaningful agent surface. Responses are subject to Chinese content restrictions on politically sensitive topics. Availability and app-store presence have been unstable in some jurisdictions.

Consumer chat app: $0, no in-app purchases on the official iOS listing. API is paid and metered per million tokens (see api-docs.deepseek.com/quick_start/pricing); exact current per-token rates were not verified for this entry. · open source

Mistral Vibe (formerly Le Chat)

Le Chat was renamed Mistral Vibe in 2026 — conversations, settings and plans carried over, and mistral.ai/vibe now resolves to the same chat surface. It has been repositioned from a chatbot into a long-horizon agent with a matching coding product (Vibe CLI, VS Code/JetBrains plugins, remote sandboxed agents, Teleport session handoff between local and remote). Every plan bundles API credits — $10/mo on free, $15–$30/mo on Pro — which makes it unusually attractive if you also build on Mistral's models. Voice input runs on Mistral's own Voxtral model, and 100+ connectors pull context from Notion, Slack, Linear, Drive and similar.

Price tiers	Free / Pro $14.99 (student $5.99) / Team $24.99 per user / Enterprise per month
Web access	Yes — web search with verified news sourcing; limited on free tier
File upload	Yes — spreadsheets, databases, PDFs, plus 100+ tool connectors
Voice	Yes — voice input powered by Mistral's Voxtral model
Platforms	Web, iOS, Android, CLI, VS Code, JetBrains, Zed

Watch out: The underlying models trail GPT-6, Claude Opus 5 and Gemini 3.8 on hard reasoning and long-context work, so you are trading capability for jurisdiction and price. The rename to Vibe collides confusingly with the general term 'vibe coding' and with the separate Vibe for Code product, making documentation hard to search. Usage is 'subject to fair usage limits' that are not published as concrete numbers. Ecosystem, third-party integrations and community content are much thinner than the US majors.

Free $0 (limited messages, web searches and coding sessions, $10/mo API credits); Pro $14.99/mo excl. tax ($5.99/mo verified students, max 12 months) with $15–$30/mo API credits; Team $24.99/user/mo excl. tax with a $50/mo minimum and 30GB storage per user; Enterprise on request.

Kimi

Moonshot AI's assistant, running Kimi K3 — a 2.8T-parameter model with a 1M-token context window aimed at coding, agentic tasks and long-horizon reasoning. Its distinguishing feature is agent swarm execution: it will dynamically deploy up to 300 sub-agents across thousands of steps for literature reviews, wide research and long industry reports. It also does lossless conversion across PPT, Word, Excel and PDF, visual slide generation from reference images, and connects to financial data sources like Wind and S&P. Paid tiers are named after musical tempos — Moderato, Allegretto, Allegro, Vivace.

Price tiers	Free / Moderato $19 / Allegretto $39 / Allegro $99 / Vivace $199 per month
Web access	Yes — web research plus data-source plugins (Wind, S&P)
File upload	Yes — PPT, Word, Excel, PDF and images, with lossless format conversion
Voice	Limited — voice input on mobile; no headline conversational voice mode
Platforms	Web, iOS, Android, desktop; open weights self-hostable

Watch out: Chinese data processing and content restrictions make it unsuitable for most Western regulated work, and the English-language product, docs and support are visibly secondary to the Chinese ones. Swarm runs consume quota fast and the relationship between tier and actual throughput is not published in concrete units. Marketing claims (300 sub-agents, 4,000 steps, AAA game generation from one prompt) run well ahead of what you should expect in practice — treat them as ceilings, not typical results.

Free tier; Moderato $19.00/mo ($180.99/yr); Allegretto $39.00/mo ($374.99/yr); Allegro $99.00/mo ($949.00/yr); Vivace $199.00/mo ($1,999.00/yr). Prices read from the official Moonshot iOS listing; kimi.com's pricing page renders client-side and could not be scraped, and CNY-denominated domestic pricing may differ. · open source

Manus

Manus is an agent rather than a chatbot: it runs its own virtual computer in the cloud, decomposes a request into a to-do list, and executes autonomously — you can close the app and it notifies you when the deliverable is ready. Typical outputs are financial models and dashboards from raw CSV/Excel, presentation decks, research reports assembled by parallel sub-agents, and small web apps. Chat is free and unlimited; agent execution is credit-metered on top of the subscription tier, with the Manus 1.6 Max architecture behind wide research.

Price tiers	Free chat / Basic $19 / Plus $39 / Pro $199 per month, plus credit packs from $20
Web access	Yes — autonomous browsing and wide research with parallel sub-agents
File upload	Yes — Excel, CSV and documents; produces spreadsheets, decks and images
Voice	Limited — mobile voice input; no conversational voice mode
Platforms	Web, iOS, Android, desktop app, Slack integration, browser operator

Watch out: Credit metering is the real cost: complex agent runs consume credits quickly, and users routinely spend well beyond their subscription tier, so treat the $19 headline as a floor rather than a budget. Long autonomous runs fail in ways that are expensive to detect — plausible-looking decks and models with wrong numbers — so every output needs review by someone who could have done the work. It is a poor interactive assistant; the asynchronous model is wrong when you want to iterate turn by turn.

Free chat; Manus Basic $19.00/mo; Plus $39.00/mo; Pro $199.00/mo; credit top-ups from $20.00 for 2,000 credits. Prices read from the official Manus iOS listing; manus.im/pricing renders client-side and its tables could not be scraped.

Gemini Notebook (formerly NotebookLM)

Google renamed NotebookLM to Gemini Notebook and moved it to notebook.google. You upload PDFs, websites, YouTube videos and text into a notebook, and the assistant answers only from those sources with inline citations — which makes hallucination far easier to catch than in open-ended chat. Beyond Q&A it generates podcast-style Audio Overviews you can interrupt and steer, Video Overviews, mind maps, flashcards, quizzes, infographics and slide decks from the same source set. It is free to use, with capacity scaling through the same Google AI Plus/Pro/Ultra ladder as Gemini.

Price tiers	Free / capacity via Google AI Plus $4.99, Pro $19.99, Ultra $99.99–$199.99 per month
Web access	Partial — can discover and import web sources, but answers are grounded only in the notebook
File upload	Yes — PDFs, websites, YouTube videos, Docs and pasted text as sources
Voice	Yes — generated Audio Overviews with interactive join-in; background and offline playback
Platforms	Web, iOS, Android; notebooks also surface in Gemini apps and Google Search AI Mode

Watch out: It is deliberately confined to your uploaded sources, so it is the wrong tool for open-ended questions or anything requiring current web knowledge. The Audio Overview format is engaging but low-density — two hosts chatting is a slow way to absorb technical material. Google has now moved the limits to an opaque compute-based system with no published per-notebook or per-day numbers, so you cannot plan capacity. The rename also broke a large amount of existing documentation and muscle memory around 'NotebookLM'.

Free $0 at standard limits; capacity scales with Google AI Plus $4.99/mo (2x standard), AI Pro $19.99/mo (4x standard) and AI Ultra $99.99–$199.99/mo (5x or 20x AI Pro). From 2 September 2026 limits are compute-based, factoring prompt complexity, models and features, refreshing every five hours up to a weekly cap. Google no longer publishes fixed per-notebook or per-day counts.

Genspark

Genspark bundles a Super Agent with purpose-built generators — AI Slides, AI Sheets, AI Docs, AI Developer, AI Meeting Notes — into a credit-metered workspace, and routes chat and media generation across third-party frontier models (GPT, Claude, Gemini, Grok for text; Nano Banana Pro, GPT Image, FLUX, Seedream for images; Veo, Sora, Kling, Seedance for video) rather than training its own. Paid tiers are sold by credit allowance rather than features: $24.99/mo for 10K credits scaling to $99.99/mo for 44K, with a $249.99/mo Pro plan above that. The pitch is one subscription instead of five.

Price tiers	Free / Plus $24.99–$99.99 by credit tier / Pro $249.99 per month
Web access	Yes — autonomous research and information gathering by the Super Agent
File upload	Yes — documents and data for sheets, docs and deck generation; GitHub integration
Voice	Yes — meeting transcription and AI Pods audio generation; voice input in-app
Platforms	Web, iOS, Android, browser extension

Watch out: Credit accounting is the whole cost model and it is opaque — the same task can cost wildly different amounts depending on which model routed it, which makes budgeting guesswork. As a reseller of other vendors' models it is exposed to their pricing and access changes, and it will always lag the source products on new features. Output quality on slides and sheets is presentable but generic, and it duplicates capability you may already own in ChatGPT or Gemini. Wrong choice if you already pay for one frontier assistant and mostly need text.

Free tier; Plus $24.99/mo (10K credits), $49.99/mo (21K), $74.99/mo (32K), $99.99/mo (44K); Plus annual $239.99; Pro $249.99/mo; credit packs $19.99 and $39.99. Prices read from the official Genspark iOS listing; genspark.ai/pricing blocks automated fetches.

Poe

Quora's aggregator gives a single interface and a shared compute-point budget across models from OpenAI, Anthropic, Google, xAI and others, plus user-created bots. It is the cheapest legitimate way to compare frontier models side by side on the same prompt without holding four subscriptions, and Basic at $4.99/mo is the lowest entry price in this category from a Western vendor. The trade-off is that you always get the model through Poe's wrapper rather than the vendor's own product surface.

Price tiers	Free / Basic $4.99 / Plus $19.99 per month (annual $49.99 / $199.99)
Web access	Partial — depends on the selected bot; not a uniform capability across models
File upload	Yes — file and image attachments, subject to per-model support
Voice	Limited — no first-class conversational voice mode comparable to the majors
Platforms	Web, iOS, Android, macOS, Windows

Watch out: You lose the first-party features that make each assistant good — ChatGPT's agents and connectors, Claude's Skills and Code, Gemini's Workspace access — and get a plain chat box instead. Compute points are consumed at very different rates per model, so a Plus plan evaporates quickly if you favour the expensive ones. New model versions arrive on Poe after they land in the vendors' own products. Wrong choice as a primary daily assistant; right choice as a cheap evaluation bench.

Free tier; Basic $4.99/mo or $49.99/yr; Plus $19.99/mo or $199.99/yr. Prices read from the official Quora/Poe iOS listing; poe.com/pricing renders client-side and the point allowances per tier could not be verified.

Proton Lumo

Proton's assistant applies the same model as Proton Mail: conversation history is stored with zero-access encryption so Proton itself cannot read it, chats are not used for training, the client code is open source, and the service is Swiss/GDPR-jurisdiction with no ad-based revenue. It does web search, image generation, file analysis and Proton Drive integration, plus custom 'Lumos' and projects on the paid tier. There is a guest mode reachable over Tor with no account at all — a genuinely unusual property in this category.

Price tiers	Free / Lumo Plus paid monthly — exact USD price not published on the product page
Web access	Yes — optional web search, off by default in keeping with the privacy model
File upload	Yes — file analysis, with Proton Drive integration
Voice	Limited — no full conversational voice mode documented
Platforms	Web, iOS, Android; Tor guest mode without an account

Watch out: The underlying model quality is well behind GPT-6, Claude Opus 5 and Gemini 3.8, and the privacy architecture is precisely what prevents the long-lived memory and deep integrations that make competitors useful. Zero-access encryption also means Proton cannot help you recover anything. Feature velocity is slow, there is no meaningful agent or coding surface, and the input box is capped around 1,800 characters. Wrong choice for heavy technical or agentic work — you are paying capability for confidentiality.

Free tier (limited Max-model access, limited messages and chat history, limited image generation, one project). Lumo Plus is a paid monthly subscription with a 30-day money-back guarantee; the exact USD price is not stated on proton.me/lumo, which shows currency-dependent pricing only — treat the number as unknown. · open source

Duck.ai

DuckDuckGo's AI chat, now at duck.ai (duckduckgo.com/aichat redirects there), proxies your prompts to third-party models with your identity and IP stripped, under agreements that the providers will not train on the conversations and will delete them within a short retention window. There is no account requirement and no login. It is the lowest-friction way to ask a capable model a question without creating an identity trail, and it sits alongside DuckDuckGo's normal private search.

Price tiers	Free — no paid Duck.ai tier confirmed
Web access	Partial — integrated with DuckDuckGo search results, not a full browsing agent
File upload	Limited — not a documented core feature; treat as unsupported for real workflows
Voice	No
Platforms	Web, DuckDuckGo browser (macOS, Windows, iOS, Android)

Watch out: Deliberately stateless: little or no memory, no file workflows to speak of, no agents, no connectors, and model choice is limited to whichever third-party models DuckDuckGo has brokered, usually mid-tier rather than flagship. Rate limits are unpublished and you will hit them. The privacy guarantee rests on contractual terms with upstream providers, not cryptography, which is a weaker assurance than Proton's. Wrong choice as a working assistant.

$0 — the page presents Duck.ai as free with no paid tier shown. DuckDuckGo has previously offered a paid higher-limit option bundled with its subscription products; no current paid Duck.ai tier was verifiable, so treat paid availability as unknown.

Qwen Studio (formerly Qwen Chat)

Alibaba's consumer assistant, now branded Qwen Studio at chat.qwen.ai and defaulting to Qwen3.7-Plus with an Auto model router. Its significance for practitioners is less the chat product than the Qwen open-weight family behind it, which is among the most-downloaded and most-fine-tuned model lineages in the world — the hosted chat is effectively a shop window you can test before self-hosting or fine-tuning the same lineage. Access is free with a login; a separate mainland-China 'Qianwen' app serves the domestic market.

Price tiers	unknown — free access with an account; paid tiers not publicly listed
Web access	Yes — web search in the chat interface
File upload	Yes — documents and images
Voice	unknown — not documented on the public product page
Platforms	Web, iOS, Android (region-dependent); open weights self-hostable

Watch out: Chinese data processing and content restrictions rule it out for most Western regulated work, and the consumer product is clearly secondary to Alibaba Cloud's commercial offering. The English interface, documentation and support are thin, the branding has churned (Qwen Chat to Qwen Studio, plus a separate domestic app), and pricing is not publicly discoverable without an account. Wrong choice as a daily driver outside China; right choice as a testbed for open weights.

unknown — chat.qwen.ai requires a login and renders its plan information client-side; no consumer subscription prices were retrievable. Free access with an account is confirmed from the product page. · open source

You.com

You.com was one of the first AI answer engines and ran a consumer assistant with multi-model chat and a Pro subscription. It has since pivoted entirely: you.com now presents only Web Search, Contents, Research and Finance Research APIs sold to developers building agents and LLM applications, benchmarked on SimpleQA, FRAMES and DeepSearchQA and priced from $5 CPM. Anyone arriving expecting a ChatGPT or Perplexity competitor will not find one; the company is now infrastructure, positioned as a cheaper substitute for built-in LLM web search.

Price tiers	Consumer plans discontinued; APIs from $5 CPM (per 1,000 search queries)
Web access	Yes — this is now the entire product, sold as an API
File upload	No — no consumer assistant remains to upload files to
Voice	No
Platforms	API only (Python SDK, REST); no consumer apps

Watch out: The consumer AI assistant this entry exists to document is effectively gone, so do not plan around it or migrate to it; existing consumer users should treat it as end-of-life. As an API it is a single-purpose search layer with no chat product, no apps and no voice, competing against providers' own bundled search. Listed here as a status marker: a first-mover in AI search that exited the consumer market.

Web Search API advertised at $5 CPM (per thousand queries); the site markets it as roughly 35% cheaper than Claude's built-in web search. No consumer assistant subscription is offered any more; per-API rate cards are on you.com/pricing. · discontinued