Quora’s Poe releases a developer API with access to a bouquet of AI models

Quora’s AI platform Poe announced on Thursday that it’s releasing an API that allows developers to easily access different models or bots for their own applications.

The API doesn’t require a separate fee. Instead, usage is tracked via Poe’s existing point-based subscription plans, where each model call costs a set number of points. For instance, low-quality image generation through GPT-4o in a 1:1 aspect ratio and 1024 x 1024 size would cost 328 points.

Today, Poe’s plans include the $4.99 per month plan (10,000 points per day), the $19.99/mo plan (1 million points per month), the $49.99/mo plan (2.5 million points per month), the $99.99/mo plan (5 million points per month), and the $249.99/mo plan (12.5 million points per month).

Developers will also be able to buy additional points at a rate of $30 for 1 million tokens. Add-on tokens don’t come in a fixed package, so customers can pay any dollar amount to get tokens based on that rate.

Through this API, developers can power tools like Cursor, Cline, Continue, Roo, and any others that work with OpenAI-compatible chat completion APIs. The platform currently provides access to more than 100 models across voice, text, image, and video generation.

These multimodal models include Imagen 4, GPT Image 1, Flux Kontext, Seedream 3.0, Veo 3, Runway Gen 4 Turbo, Kling 2.1, ElevenLabs, and Lyria.

“Currently, we are working on allowing developers to take a private bot they have built on Poe and use that through an API. Plus, we are thinking about better key management for developers for the API product,” Gareth Jones, Poe’s product lead for creators and developers, told TechCrunch over a call.

While API provides a way for developers to use different models, Poe also offers tools aimed at consumers. Earlier this year, the company introduced a new way to allow users to easily create AI-powered apps. It also offers templates to build server bots, prompt bots, and image generation bots.

At the moment, developers will need to pick and manage model use manually. Jones said that the company will consider adding budget management functionality in the future, based on developer feedback.