One API key for image and video models
Every image model takes POST /v1/images/generations and every video model POST /v1/video/generations. Both return a taskId; you poll the same path with task_id, or pass callback_url to be called when it is done. Switching models means changing the model string, plus any fields that only one model has, such as frame_count for GIFs or duration for video. Image models also answer the OpenAI SDK's images.generate in a single call.
One balance pays for everything. A request's price is set aside when it starts and charged only if it succeeds; failed requests, content-policy refusals included, cost nothing, and GET /v1/records/{id} returns exactly what one call cost. Top-ups start at $10 by card, PayPal, Alipay or USDT, and new accounts that sign up with Google (personal mailbox) or Alipay get a one-time $0.10 trial credit.
Popular models run on more than one upstream channel: when one returns an error, a rate limit or a timeout, the request moves to the next without your code doing anything. The success rates below count the last 30 days, leave out content-policy refusals and malformed requests, and are shown only for models above 98% with at least 200 calls; a dash means fewer calls or a lower rate.
Image models
| Model | Maker | Price per image | 30-day success |
|---|---|---|---|
| GPT Image 2.6 Flare | OpenAI | from $0.01 | 99.7% |
| GPT Image 2.6 Sunburst | OpenAI | from $0.01 | 99.6% |
| GPT Image 2.5 Flare | OpenAI | $0.01 - $0.3 | 99.4% |
| GPT Image 2.5 Sunburst | OpenAI | $0.01 - $0.47 | 99.7% |
| GPT Image 2 | OpenAI | $0.025+ | 99.9% |
| Nano Banana 2.1 | $0.022+ | 99.9% | |
| Gemini 3 Pro Image (Pro) | $0.068+ | no failures in 404 calls | |
| gemini-3.1-flash-image | $0.04+ | 99.8% | |
| Doubao Seedream 5.5 | Doubao | $0.03 / $0.05 | no failures in 310 calls |
| Doubao Seedream 5.0 Pro | Doubao | $0.03 / $0.06 | — |
| Qwen Image 3.0 | Alibaba | $0.035 | 99.2% |
| MiniMax H3 Image | MiniMax | $0.03 / $0.10 | — |
| Grok Imagine Image 2.0 | xAI | from $0.03 | 99.7% |
| Kling V3 Image | Kling | $0.0377 | — |
| FLUX.2 Klein 4B | Black Forest Labs | $0.006 | 99.5% |
| Muse Image | Meta | $0.02 | — |
| MAI Image 2.7 | Microsoft | ≈$0.05 / 1K image | — |
| MAI Image 2.7 Flash | Microsoft | ≈$0.015 / 1K image | — |
| Kling Image 4.0 | Kling | from $0.0628 | — |
| Kling Image 4.0 Long Image | Kling | from $0.2355 | — |
Video models
These are the best-known models; the full catalog of 183 models, including language, audio and embedding models on the same key, is on the models page.
| Model | Maker | Price | 30-day success |
|---|---|---|---|
| Seedance 2.5 | ByteDance | $0.120/s+ | 98.7% |
| Seedance 2.0 | ByteDance | $0.092/s+ | — |
| Seedance 2.0 Fast | ByteDance | $0.071/s+ | 98.4% |
| Seedance 2.0 Mini | ByteDance | $0.044/s+ | 98.1% |
| VEO 3.1 Full HD | $1.00 per video | — | |
| VEO 3.1 Fast HD | $0.08 per video | 99.6% | |
| Kling 4.0 | Kling | from $0.157/s | — |
| Kling V3 | Kling | $0.12/s+ | — |
| Wan 3.0 Video | Alibaba | from $0.045/s | 98.1% |
| MiniMax Hailuo H3 | MiniMax | from $0.097/s | — |
| MiniMax H3 Lite | MiniMax | from $0.01/s | 99.9% |
| Gemini Omni 1.2 Flash | from $0.045/s | — | |
| Grok Imagine Video 1.5 | xAI | from $0.045/s | — |
| LTX-2.3 Video | Lightricks | $0.02/s+ | — |
| Muse Video | Meta | $0.037/s | — |
When an aggregator is the wrong choice
Go to the model makers directly if you need a signed contract with them, their data-processing terms or zero data retention, batch pricing, or a public status page: APIMODELS offers none of those. Call records, prompts included, stay with your account, and requests pass through APIMODELS to its upstream channels, so confidential material belongs with a provider you have a contract with.
Quick start
cURL
# Image: same endpoint for every image model, change "model" to switch
curl -X POST https://api.apimodels.app/v1/images/generations \
-H "Authorization: Bearer $APIMODELS_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "model": "nano-banana-2-1", "prompt": "a ceramic coffee mug on a walnut desk, morning light", "aspect_ratio": "16:9", "resolution": "1K" }'
# Video: same key, same balance
curl -X POST https://api.apimodels.app/v1/video/generations \
-H "Authorization: Bearer $APIMODELS_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "model": "seedance-2.0-mini", "prompt": "steam rising from the mug as sunlight moves across the desk", "resolution": "480p", "duration": 5 }'
# Poll: GET /v1/images/generations?task_id=... or /v1/video/generations?task_id=...Python
import os, time, requests
B = "https://api.apimodels.app/v1"
H = {"Authorization": f"Bearer {os.environ['APIMODELS_API_KEY']}"}
def run(kind, body):
task_id = requests.post(f"{B}/{kind}/generations", headers=H, json=body).json()["data"]["taskId"]
while True:
d = requests.get(f"{B}/{kind}/generations", headers=H, params={"task_id": task_id}).json()["data"]
if d["state"] in ("completed", "failed"):
return d.get("resultUrls") or d.get("failMsg")
time.sleep(3)
print(run("images", {"model": "gpt-image-2.6-flare", "prompt": "a ceramic coffee mug", "aspect_ratio": "16:9", "resolution": "1K"}))
print(run("video", {"model": "seedance-2.0-mini", "prompt": "steam rising from the mug", "resolution": "480p", "duration": 5}))FAQ
How many models can one key call?
183 at the moment: 46 image, 52 video, 58 language, 25 audio and 2 embedding models, all on one balance. Some entries are the same model on a different channel or price tier, such as a Lite or a relaxed-moderation version.
Do I need accounts with OpenAI, Google or ByteDance?
No. You sign up once at apimodels.app and call every model with its key; there is no OpenAI organization verification, Google Cloud project or Volcano Engine real-name check to complete.
Is the request format the same for every model?
The endpoints, the task flow and the response shape are the same. The fields that describe the output differ where the models differ: image models share prompt, aspect_ratio, resolution and image_urls, video models add duration and their own resolution steps, and each model page lists its exact fields.
What happens when a model's upstream goes down?
If the model has a backup channel, the request moves to it automatically. If every channel fails, the request fails and you are not charged. There is no public status page, so build a retry into batch jobs rather than showing the first failure to your users.
Is there a free trial?
New accounts that sign up with Google using a personal mailbox, or with Alipay, get a one-time $0.10 credit (one per IP address in 48 hours). That is ten GPT Image 2.6 images at 1K low quality, or about two seconds of Seedance 2.0 Mini at 480p. There is no permanent free tier.
