Agnes AI
Multimodal models from Singapore Sapiens AI: one OpenAI-compatible gateway for text chat, image gen, and video. Free Access suits chatting and trying image gen without a paid Key; quotas and rate limits follow the console.
Sites and entry points
| Resource | URL |
|---|---|
| International | agnes-ai.com |
| China | agnes-ai.cn |
| API console (sign-up / Key) | platform.agnes-ai.com |
| Developer docs | agnes-ai.com/en/docs/overview |
| Model catalog (GitHub) | AgnesAI-Labs/AgnesAI-Models |
| API Base URL | https://apihub.agnes-ai.com/v1 |
Pick international or China by region; API Key and Base URL follow whatever the console / docs show now.
How to get Free Access
- Open platform.agnes-ai.com, register and sign in (usually email; whether a card is required follows the page).
- Developer console → API Keys → create and copy the Key (often starts with
sk-). - In MiniTavern use Custom (OpenAI) (chat) or Custom (OpenAI Images) (image gen) with Base URL + Key — see below.
| Item | Details |
|---|---|
| Free path | Free Access: core Flash series callable for free (RPM / fair-use limits); good for personal trials and demos |
| Paid upgrade | Token Plan: higher TPS / monthly quota for production and high-frequency agents (optional) |
| Free-tier rate limits (reference) | Text ~20 RPM; image gen by resolution (~20 RPM at 1K, stricter higher); video ~1 RPM (official Token Plan FAQ) |
| Card required | Free Access usually not; consider Token Plan for commercial use or higher limits |
Policies change. If the console says a model needs Token Plan, or you get 429, trust the site and your account permissions.
Model list
Common IDs from public docs and the catalog (case-sensitive). Flash is usually free; Pro is generally paid inference.
Text (chat / vision understanding)
Chat Completions: POST /v1/chat/completions. In MiniTavern pick Custom (OpenAI).
| Model ID | Notes |
|---|---|
agnes-2.5-flash | Recommended: chat, coding, tool use, multi-turn and image understanding; ~512K context |
agnes-2.0-flash | Previous Flash, fallback |
agnes-2.5-pro / agnes-2.5-pro-alpha | Paid reasoning (advanced coding / science, etc.) — not free |
Images (text-to-image / image-to-image)
Images API: both text-to-image and image-to-image use POST /v1/images/generations (image-to-image puts image in the body — not OpenAI /images/edits). In MiniTavern pick Custom (OpenAI Images), same Base URL https://apihub.agnes-ai.com/v1; the App routes custom protocol this way, and Test with a reference image also hits generations.
| Model ID | Notes |
|---|---|
agnes-image-2.1-flash | Recommended: dense image gen and edit, flexible sizes |
agnes-image-2.0-flash | Text-to-image / image-to-image; URL or Base64 output |
Video (text-to-video / image-to-video)
Videos: POST /v1/videos (async). Common IDs:
| Model ID | Notes |
|---|---|
agnes-video-v2.0 | Text-to-video, image-to-video, multi-image / keyframes, etc. |
agnes-video-v2.5 (or Video 2.5 in docs) | Newer video API — see official video docs |
MiniTavern has no built-in video generation entry yet; the table is for comparing site capabilities. For chat and image gen, start with Flash text / image models.
How to fill in MiniTavern
General steps: Model link.
Chat
Path: Settings → Models → Model link → New
- Platform: Custom (OpenAI)
- Base URL:
https://apihub.agnes-ai.com/v1 - Paste API Key
- Model ID:
agnes-2.5-flash(or another text model from the console list) - Test → on success, set as default
Image gen
Path: Settings → Image gen → Image gen link
- Platform: Custom (OpenAI Images)
- Same Base URL
- Same Key; model e.g.
agnes-image-2.1-flash - Test, then use
/imagineor image-gen entries
Chat and image links can differ: e.g. chat on agnes-2.5-flash, images on agnes-image-2.1-flash.
Related
- Free cloud overview — other vendors’ free / trial quotas
- Free image gen
- Model link
- Image model deploy — local A1111 / ComfyUI