An email that sounds like you. A useful study guide. A small app that actually works. Find a starting option, inspect real experiences and know what you would be paying for.
Choose a job and budget. Compare everyday assistants, real experiences, limits and costs.
A flexible starting point for questions, drafts and source-linked work, with a path from a conversation to a reviewable file or coding task.
Free entry
Usage limits apply. Regional checkout sets your final price.
Limits, available tools & source checks
New Terra reports concern Work/Codex or an API harness, not free Chat. Small patch and invoice tests show useful results alongside screen-reading and arithmetic failures. Effort, tools and checking change the cost. Inspect the product surface and keep a copy of the original input.
Discuss and revise a draft
Search for current sources
Compare supplied documents and data
Use Work or Codex for supported tool tasks
Consumer Chat and Work/Codex are separate surfaces. OpenAI says GPT-6.1 Terra is available in Work/Codex on eligible paid plans, but not in ordinary Chat; GPT-6.1 Sol is also a Work/Codex model. Check your account, workspace controls and allowance. Work/Codex share plan usage; the developer API is billed separately.
A useful candidate when the result should become an editable document, spreadsheet, presentation or small interactive artifact.
Free entry
Usage limits apply. Regional checkout sets your final price.
Limits, available tools & source checks
Selected everyday tests favoured its detailed planning; a user’s Opus 5.5 email report also shows why paid extra usage can surprise you. Neither establishes a population result. A higher effort setting or a longer answer is not automatically a better email. Keep facts and the final send under your control.
Draft and revise text
Search the web
Create downloadable files
Build an artifact or use Claude Code on eligible plans
Web, desktop and mobile. Organization settings can restrict tools or network access. Usage limits apply to free and paid plans.
Try it for a short everyday explanation or a message you want to sound natural. Test whether it follows your word limit and keeps every practical constraint.
Free entry
Usage limits apply. Regional checkout sets your final price.
Limits, available tools & source checks
A conversational style you enjoy is useful, but is not a fact-checking guarantee. The latest everyday case has unknown model/settings; the explicit Grok 4.7 review is a coding comparison. Neither supplies a fixed free-message count. Verify linked claims, quantities and any additional spend.
Chat on the web or in the official mobile apps.
Explore voice, image and file features where your account provides them.
Ask for original links when checking a current topic.
Open grok.com or its official iOS/Android app. Signing in and linking an X subscription are distinct steps. Available features, billing channel and regional checkout determine your actual access.
A useful starting point for Chinese writing, documents and everyday planning. Choose the mainland Qianwen app for its local integrations; the international Qwen product is a separate surface.
Free entry
Usage limits apply. Regional checkout sets your final price.
Limits, available tools & source checks
A 23 October newsroom inspection observed in-app shopping checkout. Offers and eligibility can change; compare the final basket and seller. This does not measure document accuracy, free PPT capacity or international Qwen. API and Qianwen Work evidence also cover different surfaces.
Draft and revise Chinese text; ask questions about supplied files.
Try the mainland app’s document, slide and work-assistant tools within your allowance.
Check shopping suggestions against actual sellers and quantities.
Mainland entry: qianwen.com and the official 千问 app. International entry: qwen.ai. Account, location, language and integrations can differ; a feature advertised for one is not promised for the other. Free entry does not make every office or creative action unlimited.
Try turning a small set of documents into an editable brief, slides or a simple prototype. Give it a concrete output and check the finished file, rather than judging only the conversation.
Free entry
Usage limits apply. Regional checkout sets your final price.
Limits, available tools & source checks
K3.1 Pro has reports of useful builds and repairs. One user reported rapid Highspeed allowance use on 23 October. On 26 October Kimi said it had adjusted settings after usage complaints; we have not verified the effect. Plans share a monthly pool and Code has a rolling limit. App and Code evidence differ; one account cannot establish your free capacity.
Ask questions about a small document set.
Try slides, documents, tables and research tools within account limits.
Check a small generated website in a real browser before sharing it.
Use kimi.com or the official Moonshot AI app. Domestic plans below are quoted in CNY; overseas offers and account eligibility can differ. Tools share an allowance, so a large research or coding run can reduce what remains for other work.
Try it for Chinese everyday questions, speaking instead of typing, a picture-based question, or a first draft. Start with a small, checkable task before using a long automated workflow.
Free entry
Usage limits apply. Regional checkout sets your final price.
Limits, available tools & source checks
The 23 October shopping inspection found dated channel-specific offers, not guaranteed savings. Verify the basket and seller. Current membership help describes model 2.1 Pro and shared advanced usage, with five-hour and seven-day limits. The game complaint is version-unknown; no latest general pass rate or fixed free count is established.
Ask by text, voice or picture in the official app.
Summarise a supplied file or turn notes into a draft.
Try creative and office tools within their separate usage limits.
Use doubao.com or the official Chinese app. This guide covers 豆包, not the separate overseas Doubao-family product or third-party sites using its name. Regional availability, account eligibility and your checkout apply.
Use the same short brief and judge clarity, tone and missed constraints. Qianwen, Kimi and Doubao are practical starting options for Chinese-language work; Grok has a recent everyday-writing test worth inspecting. This is a shortlist to try, not a performance order.
Try this: Turn the supplied event facts into a short announcement. Preserve the dates, price and unknown details. Open the ready-to-use example →
Start with the work you want to finish. Read the successes, the interruptions and the tests that kept failures visible. Each case names its product version and what was actually checked.
You want one place to explore an idea, shape a draft and turn a small set of files into useful work. It is a sensible first trial for someone who has not yet settled on one workflow.
You spend time rewriting prose, assembling a document or iterating on a small interactive tool. Test whether its editing experience reduces the work between a first draft and something you can use.
Your work already lives in Google apps. The useful question is whether editing or finding information inside that workflow saves copying and switching between tabs.
You want a free chat starting point or can define small Chinese-language extraction and coding tasks with clear checks. Developers can separately explore the API, with a spending cap and inspectable outputs.
Start free. Pay when you can name the missing feature or limit that prevents a useful result. For a normal chat subscription, the monthly charge buys an allowance; it is not a bill for each answer. API access means software calls the model and pays for usage separately.
Subscription: check the local checkout, tax, renewal date and whether the offer is monthly or an annual payment up front.
Extra usage: check paid credits, automatic top-ups, developer API charges and any third-party app fee. Set a cap before an unattended run.
Your time: include setup, checking facts, repairing outputs and waiting through limits. A cheap answer you cannot use is still unfinished work.
Will the free allowance cover my day?
There is no honest universal message count. Try one ordinary task, one correction and a follow-up on the account you would actually use. Note whether you finished, what interrupted you, the reset time and whether a paid tool was needed. Repeat on a few normal days before subscribing. A monthly pool, a rolling window and a separate tool limit can all apply at once.
A paid tier can remove a bottleneck without fixing an incorrect answer. Stop and check the source when the same factual mistake survives a correction.
Work through a small budget example
This is an invented example, not a measured product result. A $20 monthly plan used for 20 accepted tasks allocates $1 per task. If only five outputs are usable, it allocates $4 per usable result. Ten minutes of checking each of 20 tasks adds 200 minutes of your time; that time is not included in the $20.
Buying ChatGPT Plus ($20), Claude Pro on monthly billing ($20) and Google AI Pro ($19.99) together would total $59.99 per month at the checked US list prices, before any applicable taxes or extras. Test one recurring need before paying for all three.
A token is a piece of text processed by a model, not a completed task. Cached input is repeated text reused at a different price. A low bill dominated by cached input cannot price a fresh workload.
See a finished result
Three messy event notes → one useful weekend plan
Open the input, a worked answer and a tiny filterable list. Then give the same task to your own assistant.
A flexible starting point for questions, drafts and source-linked work, with a path from a conversation to a reviewable file or coding task.
Free entry Plus · US$20/month
New Terra reports concern Work/Codex or an API harness, not free Chat. Small patch and invoice tests show useful results alongside screen-reading and arithmetic failures. Effort, tools and checking change the cost. Inspect the product surface and keep a copy of the original input.
A useful candidate when the result should become an editable document, spreadsheet, presentation or small interactive artifact.
Free entry Pro · US$20/month
Selected everyday tests favoured its detailed planning; a user’s Opus 5.5 email report also shows why paid extra usage can surprise you. Neither establishes a population result. A higher effort setting or a longer answer is not automatically a better email. Keep facts and the final send under your control.
Start here when Google apps are already part of your day, or when you want to explore an image, a conversation or a Canvas project in that ecosystem.
Free entry Google AI Plus · US$4.99/month Google AI Pro · US$19.99/month
Current Argon user reports disagree sharply on allowance consumption. They are paid or unspecified accounts, not a measurement of free use. Google-app integration, stronger-model access and ordinary standalone chat have different conditions. Check the output and your own usage meter.
An accessible option to include in a no-subscription chat trial, especially if you prefer starting from Chinese-language product guidance.
Free entry
The company reported the web/app update; V4.1 tests and the low-bill story in this guide use the API. They do not establish the performance, allowance or reliability of your own chat session. Check the model if your app displays it; otherwise treat it as unknown. Verify search claims against original sources.
Try it for a short everyday explanation or a message you want to sound natural. Test whether it follows your word limit and keeps every practical constraint.
Free entry Plus · current public web listing · US$8/month
A conversational style you enjoy is useful, but is not a fact-checking guarantee. The latest everyday case has unknown model/settings; the explicit Grok 4.7 review is a coding comparison. Neither supplies a fixed free-message count. Verify linked claims, quantities and any additional spend.
A useful starting point for Chinese writing, documents and everyday planning. Choose the mainland Qianwen app for its local integrations; the international Qwen product is a separate surface.
Free entry Qianwen work assistant · Advanced · CNY19/month Qianwen work assistant · Elite · CNY49/month
A 23 October newsroom inspection observed in-app shopping checkout. Offers and eligibility can change; compare the final basket and seller. This does not measure document accuracy, free PPT capacity or international Qwen. API and Qianwen Work evidence also cover different surfaces.
Try turning a small set of documents into an editable brief, slides or a simple prototype. Give it a concrete output and check the finished file, rather than judging only the conversation.
Free entry Domestic Go · recurring monthly · CNY49/month Domestic Plus · recurring monthly · CNY99/month Domestic Pro · recurring monthly · CNY199/month
K3.1 Pro has reports of useful builds and repairs. One user reported rapid Highspeed allowance use on 23 October. On 26 October Kimi said it had adjusted settings after usage complaints; we have not verified the effect. Plans share a monthly pool and Code has a rolling limit. App and Code evidence differ; one account cannot establish your free capacity.
Try it for Chinese everyday questions, speaking instead of typing, a picture-based question, or a first draft. Start with a small, checkable task before using a long automated workflow.
Free entry Pro Standard · recurring monthly · CNY68/month Enhanced · recurring monthly · CNY200/month
The 23 October shopping inspection found dated channel-specific offers, not guaranteed savings. Verify the basket and seller. Current membership help describes model 2.1 Pro and shared advanced usage, with five-hour and seven-day limits. The game complaint is version-unknown; no latest general pass rate or fixed free count is established.
Subscription prices describe app plans, not the cost of each completed task. API use and third-party tools can have separate charges.
A developer study and a later survey show why perceived speed and finished work need separate checks. Read the tasks, dates and limits before transferring the result.
Try a working page with frozen WRS inputs. Filter dates, follow original sources and see how unknown costs are retained.
These are linked experiences with their original evidence and dates. WRS has not run a comparative account test for these products in this edition.
What supports a recommendation?
These are consumer products, not a ranking of model intelligence. Official pages support access and price facts; dated external tests and individual public accounts support narrower observations. Mainland Chinese apps can differ from international products. Recommendations are editorial starting points, not scientific grades or guaranteed results.
Official documents establish advertised features and prices. Our shortlist is an editorial judgment about a specific task. Actual account tests and user reports need their own dates, inputs and observations. A sponsor cannot buy a place in a recommendation.
Share only public or redacted examples. A suggestion is reviewed before any public case is published; there are no automatic user stars or verified-buyer badges. Icon credits and licences. Product marks identify their owners and do not imply endorsement.
WRS AI · Choose by the job
Free entry
Find my starting option
Read real experiences
