Agnes AI serves its own text, image and video models through one OpenAI-compatible API, with several Flash models currently priced at $0 and a Token Plan subscription for higher limits.
Used this tool? Rate it
Used this tool? Rate it
Agnes AI is a Singapore-based model platform that serves its own text, image and video models through a single developer API, alongside consumer apps built on the same models. The company describes itself as "a frontier AI company focused on full-modality foundation models" and says it trains its models in-house across text, image, video and reasoning. The Android app listing names the developer as Singapore Sapiens Technology Pte. Ltd., with an address at International Plaza in Singapore.
For developers the headline offer is a free AI API. Agnes AI API provides developers with unified, stable, and easy-to-integrate multimodal AI model services, supporting text, image, video, and multimodal generation and understanding capabilities. Several Flash-tier models are currently billed at $0, while the Pro text models, the full Video 2.5 model and a Token Plan subscription carry prices. Free-tier rate limits and data-use terms differ from paid plans; both are detailed below.
Agnes AI API is compatible with OpenAI-style interfaces, making it easy for developers to migrate and integrate existing projects with minimal code changes, reducing development costs and improving integration efficiency. In practice, switching an existing OpenAI-compatible API client usually means changing only the base URL, the API key and the model name.
Each model type has its own endpoint family. Select the endpoint for the model type: /chat/completions, /responses, or /messages for text; /images/generations for images; and /videos for video. Because text models answer on a Messages endpoint as well as the Chat Completions and Responses endpoints, the same key can serve both OpenAI-style and Anthropic-style clients.
Agnes Image 2.0 Flash, 2.1 Flash and 2.5 Flash share one integration contract and one price list. Agnes Image 2.5 Flash can generate images from text prompts and transform, redraw, or stylize input images, and results can be returned as image URLs or Base64 data. Output size is chosen by tier: recommended values are 1K, 2K, 3K, and 4K, combined with an aspect ratio such as 16:9. Image-to-image and multi-image composition accept one or more reference images.
Agnes Video 2.5 is an asynchronous AI video API with three generation modes: text, keyframe, or reference. Clips run from 4 to 12 seconds (5 by default) at 720P, 1080P, 1K or 2K.
The choice of mode changes what the model does with your inputs. Keyframe attempts to preserve the input image as the actual first or last frame, making it suitable for start/end composition control, whereas reference treats media as a content, style, motion or rhythm reference and may recompose or retime the result. Reference mode accepts images, audio clips and a short video, so a soundtrack can steer the pacing of the generated cut.
An earlier open-weight preview of Agnes 3.0 Flash is published on a public model hub. The Preview release has 33B parameters and a context window of 262,144 tokens. It is released under the Apache License 2.0. The model card states that the production API model uses a different checkpoint with a 1M-token context window, which does not match the 512K figure on the API documentation page; the two pages disagree, and benchmark results for the API model are not meant to be attributed to the preview weights.
Outside the API, Agnes is an agentic consumer app accessible for everyone to think, create and co-vibe together. Agnes Code is a free AI coding tool and AI code assistant that helps developers write code with AI, generate code, fix bugs, build apps, and create websites faster.
A first API call takes a few steps:
Authorization: Bearer YOUR_API_KEY with a JSON body in the OpenAI-compatible format.POST /v1/videos. Save video_id from the create response, because id and task_id identify the asynchronous task, while video_id is used to retrieve progress and results. Poll every 1–2 seconds until status becomes completed or failed.After requests start flowing, you can view your request usage, limits, and related details in the dashboard under "Usage" or "Billing". Keys from a Token Plan subscription are separate from free keys, so the key you send determines which limit pool a request draws from.
The documentation lists AI chat applications, AI search and research tools, AI writing and office productivity tools, AI image generation tools, AI video generation tools, AI character and interactive applications, AI social products, agent automation platforms, creative content production platforms, and education, entertainment, e-commerce and marketing tools as intended scenarios.
It is a weaker fit where an uptime commitment is required without a paid contract, where prompts contain data that must never be used for model training on the free tier, or for individual users under 18, as described under Limitations.
Usage-based API prices are published in USD; "list price" is the standard rate and "current price" is what is charged today.
| Model | Billing unit | List price | Current price |
|---|---|---|---|
| agnes-3.0-flash / agnes-2.5-flash | Cached input / input / output per 1M tokens | $0.005 / $0.05 / $0.15 | $0 |
| agnes-2.5-pro | Cached input / input / output per 1M tokens | $0.045 / $0.45 / $0.90 | $0.045 / $0.45 / $0.90 |
| agnes-2.5-pro-beta | Cached input / input / output per 1M tokens | $0.01 / $0.10 / $0.30 | $0.01 / $0.10 / $0.30 |
| Image 2.0 / 2.1 / 2.5 Flash | 1K / 2K / 3K / 4K per 1,000 images | $10 / $18 / $21 / $24 | $0 |
| agnes-video-2.5 | 720P / 1080P / 1K / 2K per second | $0.025 / $0.040 / $0.040 / $0.055 | Same as list |
| agnes-video-2.5-flash | 720P per second | $0.025 | $0 for a limited time |
For Video 2.5, total billable duration is the sum of the output video duration and input video duration, so a reference clip is billed at the output resolution rate; the first 5 input images are free and each further image costs $0.005. At list price, image jobs are billed by resolution tier, and the first 3 input images in image-to-image or multi-image tasks carry no extra charge. The account bill is the source of truth for actual charges, and current offers may vary by model, account eligibility, or promotion stage.
The documentation is not consistent about how long free access lasts. The FAQ says the core AI models are free to use indefinitely and can be used without a time limit, while the pricing page labels the Video 2.5 Flash offer as a limited-time promotion and keeps the full Video 2.5 and 2.5 Pro models on paid rates.
| Plan | Monthly price | First month | agnes-3.0-flash quota | Image quota | Video quota |
|---|---|---|---|---|---|
| Starter | $4 | $2 | 1,500 requests per 5 hours; 15,000 per week | 4,000 images per day | 500 seconds per day |
| Plus | $10 | $5 | 7,500 requests per 5 hours; 75,000 per week | 4,000 images per day | 500 seconds per day |
| Pro | $50 | $25 | 30,000 requests per 5 hours; 300,000 per week | 4,000 images per day | 500 seconds per day |
Annual billing is priced at ten months of the monthly rate, shown as "2 months free", and the first-month discount applies to monthly plans. The plan card states that text requests are powered by Agnes-3.0-Flash, usually about 100 TPS, 150 TPS during off-peak hours. Rate limits and quota windows apply at the same time, so a Pro key with 1000 RPM for text models is still capped by its 5-hour and weekly totals.
Agnes AI differs from both in serving only its own model family, several of them currently at $0.
Agnes 3.0 Flash, Agnes 2.5 Flash, the three image models and Video 2.5 Flash currently cost $0. Agnes 2.5 Pro, 2.5 Pro Beta and Video 2.5 are billed, and the free tier is limited to 10 text requests per minute.
Yes, in most cases. Point the client at https://apihub.agnes-ai.com/v1, use an Agnes API key and change the model name; video uses a separate asynchronous task flow.
On the free tier, prompts and outputs may be used to improve models unless you opt out. Paid Token Plan and enterprise content is excluded by default. The Terms of Service describe a broader opt-out license without that tier split.
The Agnes AI Token Plan starts at $4 a month for Starter, with Plus at $10 and Pro at $50, each 50% off in the first month; annual billing costs ten months' worth.
Video 2.5 produces 4–12 second clips at 720P, 1080P, 1K or 2K; Video 2.5 Flash is limited to 720P.
App store listings name Singapore Sapiens Technology Pte. Ltd. as the developer, and the Terms of Service are governed by Singapore law.