Poe AI
Poe puts GPT, Claude, Gemini, Grok and 100+ models under one subscription, plus a no-code custom bot builder. Billed in compute points that vary wildly by model.
The widest model selection in one app, undercut by a compute-points system that hides the price of a message until after you have paid it.
Should you use Poe?
Model-hoppers and comparison shoppers who want GPT, Claude, Gemini and Grok under one login and build their own custom bots.
You live inside one model's ecosystem. Native Custom GPTs, Claude Projects and provider tools are missing here.
The free tier is 300 points a day, which one strong answer can wipe out. Steady use starts at $49.99/year (10k points/day).
No message shows its cost before you send it. The same prompt cost me 24 points on one bot and 276 on another.
What is Poe?
Poe does not run its own AI model. It is a counter where every other company's models are on sale under one subscription, so you can send a prompt to OpenAI's GPT line, switch to Anthropic's Claude, then to Google's Gemini or xAI's Grok without opening four tabs or paying four bills. Owned by Quora and live since 2023, it carries well over a hundred official models plus a marketplace of community bots that runs into the millions, and it works on web, iOS, Android and desktop.
The catch sits in how you pay. Instead of a flat message allowance, Poe runs on compute points, and every model burns them at a different rate that you cannot see until the answer is already spent. That design is the dividing line in every review of this product: people who want breadth under one login rate it highly, and people who get surprised by a 2,971-point charge for a single explainer rate it near the floor.
Key features & how they perform
Each feature rated from hands-on testing and aggregated review sentiment.
100+ models in one box
GPT, Claude, Gemini, Grok, DeepSeek and image and video generators, all reachable from a single prompt field with no extra logins.
Side-by-side comparison
Run the same question through several models in one thread. The feature is excellent, but each comparison spends points, so it is the fastest way to empty a balance.
Custom bot builder
Pick a base model, write a system prompt, publish. No code, free to set up, and shareable with the whole Poe audience.
Image & video generation
Dedicated media bots cover image creation and editing plus text-to-video. Capable, but these are the most point-hungry requests on the platform.
Compute-points billing
The single most criticized part of the product. Costs swing wildly between models and are never shown before you hit send.
Cross-device sync
Chats follow you across web, mobile and desktop. The interface is clean and the apps are stable, which reviewers consistently note.
Feature ratings blended from Trustpilot, G2, App Store, Play Store & Reddit review patterns + hands-on testing.
I spent a day on Poe's free plan and watched it run dry
What 300 daily points actually buy, from the first answer to the first locked door.
What the free plan actually hands you
The free account opens with 300 compute points for the day and a countdown to when they refill. Poe prints a dollar value beside the balance, which came to about $0.0091.
That single pool has to cover everything: chat, image work, and any model comparisons. There is no separate allowance per feature.
Three hundred reads as a lot until you learn a single answer from a strong model can cost more than ten times it. That gap is the whole story of the free tier.
Picking a model is the part Poe gets right
The one feature that clearly earns Poe its reputation is model choice. From a single box I could aim my question at Assistant, GPT-6-Astra, Claude-Opus-5.5 or Claude-Fable-5.1, with more sitting behind the search icon.
No separate logins, no juggling tabs. This is the reason to use Poe over any single chatbot.
For a first test I sent a practical question to the Assistant bot, the house default that tends to be one of the cheaper options.
"Give me 5 tips for filing a US federal tax return as a freelancer."
The first wall came faster than I expected
Under that answer, Poe offered to run the same question through other models for comparison. I tapped Compare on Grok-4.20-Multi-Agent.
Instead of a second opinion, I got a paywall. The popup wanted 665 points for that one comparison, and I had 276 left.
The comparison simply did not run. Grok returned a blocked state with a short line saying it needed more points to answer.
Building your own bot is free and simple
Custom bots are another genuine strength. The Create screen lets you build a prompt bot, an image or video generation bot, a role play bot, a script bot or a server bot, plus a Canvas option for a richer web experience.
I built a prompt bot on top of GPT-5.4 to see how much the builder asks of you.
"You are a friendly US small-business tax helper. Answer in plain English with dollar examples."
One normal request costs more than ten days of free points
I opened a fresh chat and asked GPT-6-Astra for something completely ordinary, the kind of request anyone would make on day one.
"Write a detailed 800-word explainer comparing Roth IRA vs Traditional IRA for a US reader, with examples."
The request needed 2,971 points. My balance was 276. Put plainly, one normal explainer costs nearly ten full days of free points before a free user could even afford to send it.
Image editing was no kinder. I attached an AI-generated photo so nobody real is in it, and asked GPT-Image-2.5-Sunburst for one small change.
"change the shirt color"
That one edit wanted 2,853 points against the same 276 balance. Same story, same locked door.
The same prompt, wildly different prices
The clearest proof of how unpredictable the pricing is came from running my freelancer tax prompt again, this time through GPT-5-mini.
The answer itself was excellent. It used a visible reasoning step and cited irs.gov next to each tip, pointing at Schedule C, Publication 334 and Form 1040-ES.
Then I checked the activity log in USD and points, and the picture snapped into focus. The identical freelancer question cost 24 points on Assistant and 276 points on GPT-5-mini.
Poe pricing
Figures taken from the official Poe subscription page. All prices shown are billed yearly.
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | 300 compute points/day · all models usable until points run out · custom bot creation |
| 10k points/day Popular | $49.99 /yr | 10,000 points refreshed daily · works out to $4.17/mo · entry point for steady personal use |
| 660k points/month | $199.99 /yr | 660,000 points/month · $16.67/mo · pooled monthly instead of daily |
| 1.65M points/month | $499.99 /yr | 1.65 million points/month · $41.67/mo · heavy multi-model or image use |
| 3.3M points/month | $999.99 /yr | 3.3 million points/month · $83.33/mo · power users and small teams |
| 8.25M points/month | $2,499.99 /yr | 8.25 million points/month · $208.33/mo · top tier for the heaviest workloads |
Poe shows yearly pricing by default, with a monthly toggle. The yearly view claims a 17% saving over paying monthly.
Pros & cons
Specific conclusions from testing and real user reviews, not generic filler.
✅ Pros
- GPT, Claude, Gemini, Grok and DeepSeek reachable from one prompt field
- Same prompt runs across several models in a single thread
- No-code bot builder publishes to millions of Poe users
- Bot creators earn revenue when others use their bots
- Clean interface and stable apps across web, mobile and desktop
- Avoids being locked to one AI company's roadmap or outages
⛔ Cons
- No message shows its point cost before you send it
- Identical prompts cost ten times more on one model than another
- Missing Custom GPTs, Claude Projects and provider-native tools
- Custom bots fed large PDFs drift and hallucinate, users report
- Prompt bots cannot take external actions without the server API
- You stay one layer behind each model's newest features
Synthesized from real reviews on Trustpilot, G2, Reddit, the App Store & Play Store · paraphrased, not quoted
Poe scorecard
Rated against what a multi-model AI aggregator is actually built to do.
| Dimension | Verdict | Score |
|---|---|---|
| Model selection & access Range of models under one login | Excellent |
9.5
|
| Custom bot creation No-code builder and marketplace | Excellent |
8.5
|
| Model comparison Running one prompt across models | Good |
8.0
|
| Answer quality Usefulness of the underlying models | Good |
8.0
|
| Ease of use Signup, navigation, model switching | Good |
7.5
|
| App stability & sync Performance across devices | Good |
7.5
|
| Pricing transparency Knowing a cost before you pay it | Weak |
3.5
|
| Value for money Work done per dollar spent | Weak |
5.0
|
| Free tier usability What 300 daily points achieve | Weak |
4.0
|
| Depth vs native apps Access to provider-only features | Average |
6.0
|
| Scrutool Score Equal-weight average of all 10 dimensions |
6.8
|
What users say about Poe
The themes reviewers raise most often, by share of analysed reviews.
% = share of analysed reviews mentioning each theme (Trustpilot, G2, App Store, Play Store, Reddit)
The best model buffet in AI, served with a bill you cannot read until you have eaten
Poe is the easiest way to reach every major AI model from one place, and the custom bot builder and comparison view are genuinely good. Nothing else matches its breadth under a single login. The problem is the compute-points system, and it is not a minor one. On the free plan I burned through the daily allowance with a single good answer, watched the same prompt cost 24 points on one bot and 276 on another, and never once saw what a message would cost before sending it. That unpredictability is the loudest theme in real reviews, not a quirk I happened to hit. People who want to compare models, build bots and consolidate several subscriptions into one bill will find real value, especially on the $49.99 yearly tier. Anyone who leans on a single model and wants its native tools, or who needs to know what each message costs, is better served going direct to that provider.
Discussion
Join the discussion and share your perspective.