Image pricing is quoted in two incompatible ways: a flat price per image, or a token price where each image burns a different number of tokens depending on its size. This table converts both into one generated image, so the numbers can be compared directly.
| Model | Provider | Billing | Per image |
|---|---|---|---|
| CogView3-Flash | Zhipu AI | free | FREE |
| Wan 2.2 T2I Flash | Alibaba Cloud | flat | $0.0215 |
| Seedream 4.0 | ByteDance | flat | $0.0270 |
| Seedream 5.0 | ByteDance | flat | $0.0270 |
| Wan 2.5 T2I Preview | Alibaba Cloud | flat | $0.0308 |
| Hunyuan Image 3.0 | Tencent | token (20K) | ~$0.0308 |
| Seedream 4.5 | ByteDance | flat | $0.0340 |
| Seedream 5.0 Pro | ByteDance | flat | $0.0340 |
| Seedream v5.0 Lite (TokenHub) | ByteDance | token (22K) | ~$0.0338 |
| WAND Vega Image Lite | Tencent | token (22.5K) | ~$0.0346 |
| Seedream v5.0 Pro (TokenHub) | ByteDance | token (60K) | ~$0.0923 |
| Vidu Image Q2 | Vidu | token (93.75K) | ~$0.1442 |
| WAND Vega Image Flash | Tencent | token (100.8K) | ~$0.1551 |
| WAND Vega Image Pro | Tencent | token (171K) | ~$0.2631 |
| Volume | Cheapest paid | Seedream 5.0 | WAND Vega Pro |
|---|---|---|---|
| 100 images | $2.15 | $2.70 | $26.31 |
| 1,000 images | $21.50 | $27.00 | $263.10 |
| 10,000 images | $215 | $270 | $2,631 |
For product shots and social assets, Seedream-class models at $0.027 cover most needs. The 10x jump to WAND Vega Pro is only worth it when the image is the product — hero images, print, or assets that will be retouched anyway.
CogView3-Flash costs nothing on NovAI: 10 free-model calls/day and 5 requests/minute on the free tier. It is enough to wire up generation, storage and CDN delivery before spending on the good models.
curl https://aiapi-pro.com/v1/images/generations \
-H "Authorization: Bearer $NOVAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"cogview-3-flash","prompt":"a paper boat in the rain, macro"}'Three factors: output resolution and quality tier (lite vs pro), the provider's compute cost, and the billing model — some models charge a flat price per image, others bill per token and the image consumes 20K-171K tokens depending on its size.
Yes. CogView3-Flash is free to call with a NovAI API key; the free tier allows 10 free-model calls per day and 5 requests per minute.
They are the providers' list prices passed through with a 1.0 multiplier and zero platform markup, published at /china-ai-leaderboard.json so they can be checked.
At the cheapest paid rate (Wan 2.2 T2I Flash, $0.0215/image) about $21.50; with Seedream 5.0 at $0.027 it is $27; with WAND Vega Pro at $0.263 it is $263. The free tier covers testing only.
111 live models, provider list price, 0% platform markup. $2 credit on signup, no card required.
Get an API keyPrices reflect provider rate cards as of 2026-10-22 and are published machine-readably at /china-ai-leaderboard.json. Token-billed figures are estimates based on each model's tokens-per-image; flat figures are the published rates.