Poe Review, Pricing & Alternatives (2026)

Poe is Quora’s consumer and prosumer service for hosted AI models, bots, and apps. Quora launched it in 2023, and Poe aggregates models from providers including OpenAI, Anthropic, Google, Meta, xAI, Kimi, DeepSeek, and FireworksAI. The service is an access layer rather than a foundation-model company. Poe does not give subscribers downloadable weights, and an open or open-weight model in its catalog does not make Poe open source. Its audience includes writers comparing responses across providers, people generating media, and creators publishing bots or apps. Poe also supports creator monetization. Consumer plans combine daily or monthly points with access to a catalog that can change as providers and models are added or removed.

A typical Poe session starts with choosing a bot, model, or app, then entering a prompt and spending points according to that tool’s rate. Users can compare responses across providers, continue conversations with supported context, generate images, video, or audio, and run creator-built tools that combine models. Poe Apps, introduced in February 2025, combines models with JavaScript and custom logic. Script Bots, introduced in November 2025, combine more than 200 models and run Python in Poe’s secure environment. Creators can publish bots and apps, while subscribers use them through Poe’s hosted interface. The public-bot API, introduced in August 2025, extends this arrangement to developer applications through compatible API formats and Poe-funded point balances.

Poe’s technology is an orchestration layer over third-party models, not one named foundation model. Its catalog spans GPT, Claude, Gemini, Grok, Kimi, DeepSeek, image models, video models, and audio models, although availability changes. Recent additions include Poe Apps in February 2025, GPT Image Generation and GPT-Image-1 in April 2025, the public-bot API in August 2025, Script Bots in November 2025, Opus 4.5 defaults and FLUX-schnell in December 2025, and multiple API keys in March 2026. The API supports OpenAI-compatible access, while current creator materials also describe Anthropic-compatible access. These calls use subscriptions or add-on points. Poe therefore provides hosted model access, rather than weights, local inference, or a separate model catalog that developers control.

The free plan costs $0 and provides daily points that reset every 24 hours without rolling over. Paid plans publicly documented by Poe cost $4.99/month for 10,000 points per day, $19.99/month for 1 million points per month, $49.99/month for 2.5 million, $99.99/month for 5 million, and $249.99/month for 12.5 million. Add-on points cost $30 per 1 million, expire after one year, and are non-refundable and non-transferable. Per-bot rates vary, so no universal message quota exists. Paid access can add higher limits, more models and media, and supported-bot context up to 2 million tokens. Taxes may apply at checkout. Exact annual USD prices are not exposed in the referenced pricing text.

Poe’s main limitation is dependence on changing provider access, bot rates, privacy settings, and point economics. ChatGPT offers a focused OpenAI service, Claude a focused Anthropic workspace, and Gemini a Google-centered product, while Poe trades provider-specific depth for access to many models in one account. Official bots generally do not train on chats, but half-shield third-party bots and apps may share data or use interactions for training. Poe also does not provide model weights, self-hosting, or a blanket commercial redistribution license. The public-bot API extends hosted access, not ownership of underlying models. Pick Poe if you want one writing workspace for comparing models and media tools without managing separate provider accounts.