Eleven image models, list prices in USD, normalised to the unit a batch is actually budgeted in: cost per 1,000 images. Spread: 6.8x.
| Model | Per image | 1,000 images | Images per $10 |
|---|---|---|---|
| Wan 2.2 T2I Flash | $0.0215 | $22 | 465 |
| WAND Vega Image Lite | $0.0250 | $25 | 400 |
| Seedream 5.0 | $0.0270 | $27 | 370 |
| Vidu Image Q2 | $0.0290 | $29 | 345 |
| Wan 2.5 T2I Preview | $0.0308 | $31 | 325 |
| Hunyuan Image 3.0 | $0.0310 | $31 | 323 |
| Seedream v5.0 Lite (TokenHub) | $0.0340 | $34 | 294 |
| Seedream 5.0 Pro | $0.0340 | $34 | 294 |
| Seedream v5.0 Pro (TokenHub) | $0.0460 | $46 | 217 |
| WAND Vega Image Flash | $0.0690 | $69 | 145 |
| WAND Vega Image Pro | $0.1460 | $146 | 68 |
List prices checked 2026-10-30 from the live price feed. These models bill per image, not per megapixel, so on most rows a 4K render costs the same as a 1K one — check the model page before assuming otherwise.
1. Reject rate is the hidden multiplier. At a 30% discard rate you generate 1.4 images per usable one. 1,000 usable images is 1,400 generations: $30 on the cheapest tier, $204 on the most expensive.
2. The cheapest model is not the cheapest pipeline. If a model needs more retries for a usable result, the nominal price advantage disappears. Measure your own hit rate per model before switching.
3. Free options exist for the first hundred. CogView3-Flash and GLM-4.1V-Thinking-Flash are $0 on the free tier (10 calls/day, no card) — enough to wire the pipeline before paying.
Provider list price, 0% platform fee, free tier without a card.
Get a free API key Batch calculator