Get the most out of your AI: Grok, Claude and 4 more apps
Updated and share it with people you trust (18+. Features and limits from each company's own pages (linked); tips and example prompts are ours.
The features worth turning on, the settings that matter, things to try today, and the mistakes that burn through your limits.
Grok
Newest Grok model on Arena Text right now: Grok 4.7 (#96). See its strengths and limits on its page. The older Grok 4.5 still ranks higher (#52).
Features worth turning on
- Voice: talk with Grok out loud (Plus and up)
- Grok Imagine: images, and video on Premium and up
- Private Chat: not saved to your history and not used for training
- Grok Bot (Premium and up): an agent with its own cloud computer
- Grok Bot phone apps (rolling out on iOS and Android): let Grok Bot use Calendar, Reminders, Contacts and more, allowing each app separately
- Search of public X posts and the web
Settings that matter
Mode menu: Fast for quick replies; Expert for hard questions; Heavy for the toughest work (paid tiers). How to choose a model and effort level.
Try these
Prompt: what X is saying about an event
Prompt: turn a pet photo into a poster
Prompt: explain a chart
Got X Subscription Premium or higher? It includes Grok Bot, a team of AI helpers with their own cloud computer. What Grok Bot can actually do for you.
Mistakes that waste your limits
- Forgetting the pool is shared: Grok Chat, Imagine, Voice, Bot and Cursor all draw from one weekly allowance, and xAI says usage costs vary by product
- Using Heavy for simple questions
- Not checking remaining usage in settings before a big job
2,3,4,5 Plans and limits: Grok (xAI) plans, strengths and limits.
Claude
Top Claude model on Arena Text right now: Claude Opus 5.5 (#2). See its strengths and limits on its page.
Features worth turning on
- Projects: a workspace with your files and instructions (up to 5 on Free)
- Memory across conversations
- Research (Pro and up): Claude searches and writes a report
- Connectors: link apps and tools you use
- Artifacts and file creation: documents, slides and small apps made in the chat
- Claude Code (Pro and up) for coding; Claude in Chrome for browser tasks
- Phone app: stay signed in to a work and a personal account and switch without signing out; voice mode in Apple CarPlay (iPhone)
Settings that matter
Model menu next to the send button: Sonnet for most things, Opus (Pro) for hard work; Effort Low/Medium stretches your usage, High is the best balance. How to choose a model and effort level.
Try these
Prompt: fix the weakest lines in a cover letter
Prompt: a printable moving checklist
Prompt: flag unusual terms in a contract
New on your phone: the Claude app can keep a work and a personal account signed in at once, and voice mode works in Apple CarPlay. What changed and how to get it.
On Claude Max? Max and Ultra include US$100 or US$200 a month in API credits, separate from your chat limits. How to claim them and 5 things to do with them.
Running out of Claude? Get more out of your Claude plan for less: fresh chats, Projects, model vs effort, and the Claude Code commands that save the most.
Mistakes that waste your limits
- Very long chats: Anthropic says longer, more complex conversations use more of your limit, so start a new chat for a new topic
- Forgetting that the web, desktop, phone apps and Claude Code share one usage pool
- Leaving effort on Max for routine tasks
- Not checking Settings > Usage to see where you stand
6,7,8 Plans and limits: Anthropic plans, strengths and limits.
Gemini
Top Google AI model on Arena Text right now: Gemini 4 Argon (#1). See its strengths and limits on its page.
Features worth turning on
- Gems: saved custom helpers for jobs you repeat
- Deep Research: a report from hundreds of sources
- Gemini Live: talk out loud
- Canvas: write and edit documents or code alongside the chat
- Files in chat: have Gemini make a PDF
Settings that matter
Doc How to choose a model and effort level.
Try these
Prompt: Sheet or slide deck
Gems become skills on Nov 17, 2026
On personal Google accounts (18+), Gems turn into skills and go read-only that day, and they're removed on Dec 17, 2026. Make new ones as skills (Customize › Skills). Google moves existing Gems over for you, but audio, video, old Office (.doc, .xls, .ppt) and zip files don't carry over, so download them before Dec 17. Work accounts follow in March 2027, school accounts in June 2027. 1
New, on every plan: you can now edit the files Gemini makes right in the chat, or highlight one part for Gemini to rewrite. How it works and what works on your phone.
Sharing a Gemini avatar? Friends can put your face and voice in their images and videos once they accept. Removing them later doesn't delete what they already made. What to know first.
Mistakes that waste your limits
- then fix it by hand or highlight one part for Gemini to rewrite (every plan; PDFs on phone too
Sources: , , checked . Plans and limits: Google plans, strengths and limits.
ChatGPT
Top ChatGPT model on Arena Text right now: GPT-6.1 Sol (#27). See its strengths and limits on its page.
Features worth turning on
- Projects: keep files and chats for one topic together
- Memory: ChatGPT can remember details between chats (limited on Free; Go and up add memory of past chats)
- Deep research: a cited report from many web pages (limited on Free and Go)
- Scheduled tasks: run a prompt at a set time, like a daily briefing (3 active on Free and Go, 5 on Plus, 15 on Pro)
- Study mode: walks you through a topic step by step
- Voice (with video on Go and up)
Settings that matter
Pick GPT-6 and the Thinking level that fits the job (Instant for quick questions; Medium/High for hard ones). Free and Go: use Think for harder questions. How to choose a model and effort level.
Try these
Prompt: plan my week
Prompt: quiz me on a topic
Prompt: a daily weather and news summary
Claimed the US student offer? It ended Oct 31, 2026. How to avoid the US$20 charge when your free months end.
On ChatGPT Pro? Since Oct 29, 2026, Pro 100 and Pro 500 get twice their Work and Codex allowance, and the planned Pro 200 cut is canceled. What changed.
On ChatGPT Pro or Business? Your plan now includes monthly API credits (US$100 to US$500 on Pro) that you have to claim. How to claim them.
Mistakes that waste your limits
- Pasting whole documents when only a section matters: OpenAI's own tips for ChatGPT Work and Codex say limiting source material makes your limits last longer
- Asking for long answers you don't need: set the length and format up front
- Making lots of images: in ChatGPT Work and Codex, OpenAI says image generation uses limits about 3–5x faster
- Using the highest thinking level for simple questions
9,10,11 Plans and limits: OpenAI plans, strengths and limits.
Copilot
Features worth turning on
- Smart mode: picks quick or deep thinking for you
- Think Deeper and Study and learn modes
- Copilot in Word, Excel, PowerPoint and Outlook (Microsoft 365 Personal, Family, Premium)
- Memory: ask “What do you know about me?”, or say “Remember that…” / “Forget that…”
- Image creation
Settings that matter
Mode menu at the bottom left: Smart is the safe default; Think Deeper for complex questions; Quick response for speed. How to choose a model and effort level.
Try these
Prompt: meeting notes to a summary in Word
Prompt: explain and fix an Excel formula
Prompt: remember a preference
Mistakes that waste your limits
- Expecting Copilot in Office on the free version: it needs a Microsoft 365 Personal, Family or Premium subscription, and only the subscription owner gets it
- Using Think Deeper for quick questions
- Not telling Copilot what to remember (or forget)
12,13,14 Plans and limits: Microsoft · Copilot plans, strengths and limits.
Perplexity
Features worth turning on
- Search with sources on every answer
- Deep Research: in-depth reports with charts
- Learn Mode: interactive explanations and quizzes
- Model choice and Thinking (Pro and up)
- Model Council (Max): ask several models at once
Settings that matter
Leave the model on Best for everyday searches; on paid plans pick a model and turn on Thinking for multi-step questions. How to choose a model and effort level.
Try these
Prompt: what changed in a Canadian rule
Prompt: compare products with sources
Prompt: learn a topic, then get quizzed
Mistakes that waste your limits
- Using Deep Research for a quick fact: Best is faster and doesn't use your Pro Search allowance
- Not opening the sources: check the links for anything important
- Switching models for every question: Perplexity suggests starting with Best
15 Plans and limits: Perplexity plans, strengths and limits.
Tips that work in every app
- Start a new chat for a new topic. Long chats use more of your limit and can confuse the AI.
- Say who it's for, how long, and what format. "Three bullet points for my boss" beats "summarize this".
- Give it the source. Paste the email, upload the document, or attach a photo instead of describing it.
- Ask it to check itself. "What might be wrong with this answer?" often catches mistakes.
- Check anything important against a real source.
Ready to let AI do tasks for you?
- AI agents, explained for normal people
- How to run a small team of AI agents (without being a programmer)
- Get more out of your Claude plan for less
- What Grok Bot can actually do for you
Sources (15)
- Gemini Apps Help: About the transition from Gems to skills
- x.ai/pricing
- xAI: Consumer FAQs
- grok.com
- x.com/bot/status/2113393684209176718
- claude.com/pricing
- Claude Help Center: Model, effort and thinking settings
- x.com/claudeai/status/2115859377488072879
- learn.chatgpt.com: pricing
- OpenAI Help: GPT-6 and other models in ChatGPT
- chatgpt.com/pricing
- Microsoft Support: Copilot conversation modes
- Microsoft Support: Copilot privacy controls
- microsoft.com/en-us/microsoft-365-copilot/pricing/individ…
- perplexity.ai/help-center/en/articles/10354919-what-advan…
Sources checked Oct 22 to Oct 30, 2026.