Skip to content

Poe

Quora's app for chatting with many AI models (GPT, Claude, Gemini and more) and user-made bots under one subscription.

Reviewed by Andrew Miller · Updated

Visit Poe Free plan

Our rating

3.6/5

A useful switchboard for sampling many models on one bill, but your chats travel to every provider and bot developer you touch, and points obscure the real price of each message.

What it is

Poe, made by Quora, puts models from OpenAI, Anthropic, Google and many others in one app, along with image, video and audio models and bots built by users. You can switch models mid-conversation, and a memory feature carries context between bots from different providers.

Usage is measured in compute points. Each plan includes an allowance, stronger models cost more points per message, and add-on points can be bought when it runs out. The same points pay for Poe's OpenAI-compatible API, which reaches hundreds of models and public bots through one key at up to 500 requests a minute.

The cheapest paid plan costs $4.99 a month, added in 2025; the main plan is $19.99 a month.

Our review

Poe earns its place when breadth matters more than depth. One subscription covers GPT, Claude, Gemini, open models and image and video generators, and the $4.99 entry plan makes it a cheap way to sample paid models. Developers get the same catalogue through an OpenAI-compatible endpoint, so moving a prototype from one provider to another is a change of model name rather than a new SDK and billing account.

The cost model is opaque by design. Everything is priced in compute points, top models burn far more per message than cheap ones, and Poe does not publish a cash value for the points in a subscription. For add-on points, its developer docs say the aim is to charge what the underlying provider charges, so Poe is not a discount route for heavy use. When the balance hits zero the API returns a 402 error rather than slowing down.

Privacy is the part most buyers miss. Poe's privacy policy, updated on 30 April 2026, says model providers and third-party bot developers receive the contents of your chats and uploads "to provide and generally improve their services", and that developers who build bots on Poe may store chats on their own servers to train their models. The memory feature shares context across providers as well.

You also give up what model makers keep for their own apps: OpenAI's Codex, Claude Code, Gemini's Google Workspace links and the providers' newest agent tools. The API has gaps of its own, with private bots unsupported, audio input dropped, and parameters such as seed and response_format silently ignored. Poe is a front end to many models, and the more you rely on one of them, the more the front end gets in the way.

Where it earns its place

  • A solo developer prototyping a customer-facing chatbot

    Compare Claude, GPT and Gemini on the same prompts before choosing a provider

    Point the OpenAI SDK at api.poe.com/v1 with a Poe key and swap model names; calls draw on subscription points, then add-on points priced to match provider rates. Expect a 500-requests-a-minute cap and 402 errors when the balance runs out.

  • A marketing manager who writes copy and makes images each week

    Use several text and image models without four subscriptions

    The $19.99 plan covers occasional use of many models in one place. Daily heavy use of one premium model drains points quickly, at which point that provider's own plan gives more usage and its own features.

  • A compliance officer at a 50-lawyer firm

    Decide whether staff can use Poe for client work

    Poe's privacy policy says chat contents and uploaded files go to the model provider and, for third-party bots, to developers who may store and train on them. That is hard to square with client confidentiality, so client material should stay out of Poe.

Before you commit

Setup
Sign up on the web or in the apps and pick a model; API keys are created at poe.com/api/keys and work with the OpenAI SDKs by changing the base URL.
Integrations
OpenAI-compatible Chat Completions and Responses endpoints covering text, image, video and audio models and public bots; private bots are not available through the API.
Data and privacy
Chat contents and uploads go to the model provider and, for third-party bots, to developers who may store and train on them; memory is shared across providers. The privacy policy names no security certifications.
Cost at scale
Subscription points cover chat and API alike; beyond the allowance, add-on points are priced to match provider rates, so a production workload pays roughly list price with no published volume discount.
Limits
500 API requests a minute, counted by request rather than token; n must be 1; seed and response_format are ignored; a zero balance returns 402.

What's good

  • GPT, Claude, Gemini and media models under one bill
  • Cheap $4.99 entry plan
  • OpenAI-compatible API that draws on subscription points
  • Memory carries context across providers
  • User-built bots with creator payouts

What's not

  • Chats and uploads shared with model providers and bot developers
  • No published cash value for subscription points
  • API ignores seed and response_format and drops audio input
  • Lacks the providers' own coding agents and integrations

Best for

People who want to try and compare many AI models without separate subscriptions.

Alternatives

Poe: common questions

Is Poe worth it?
A useful switchboard for sampling many models on one bill, but your chats travel to every provider and bot developer you touch, and points obscure the real price of each message. We rate it 3.6 out of 5.
How much does Poe cost?
Free tier with a limited points allowance; paid plans from $4.99/month ($49.99/year), with the main plan at $19.99/month ($199.99/year) on the US App Store. Add-on points available; chat and API usage both metered in compute points.
What are the downsides of Poe?
Chats and uploads shared with model providers and bot developers; no published cash value for subscription points; aPI ignores seed and response_format and drops audio input; lacks the providers' own coding agents and integrations.
Who is Poe best for?
People who want to try and compare many AI models without separate subscriptions.
How much does Poe cost for a team or in production?
Subscription points cover chat and API alike; beyond the allowance, add-on points are priced to match provider rates, so a production workload pays roughly list price with no published volume discount.
Is Poe safe for company data?
Chat contents and uploads go to the model provider and, for third-party bots, to developers who may store and train on them; memory is shared across providers. The privacy policy names no security certifications.
What are the best Poe alternatives?
The closest alternatives we have reviewed are ChatGPT, Claude and Perplexity.

Answers checked Oct 23, 2026.