Wholesale prices on AI image generation — one key, pay-as-you-go
Five image models at posted, per-image prices — $0.007 to $0.009 per image, and the price stays the same at 1K, 2K and 4K. Plus the top text and coding models on the same OpenAI-compatible key. No subscription, no minimums, cancel any time.
Get your API key → · See all prices · Live prices (sign-in required) · Latency & status
Works with your stack
One base URL, one key, OpenAI-compatible — swap models with a string, no re-platforming:
Text & coding: GPT-6.1 · GPT-6 (Sol / Astra / Luna) · GPT-5.6 (Sol / Terra / Luna) · Claude Opus 5.5 / 4.8 · Claude Sonnet 5.5 / 5 · Gemini 3.8 Flash · Grok 4.7 / 4.5 · GLM 5.3 · Kimi K3 Image: gpt-image-2 · gpt-image-2.5 · gemini-3.0-pro-image-ultra · gemini-3.1-flash-image-ultra · gemini-nano-banana-2.1-ultra Video: not on sale right now — see Video Generation
Claude also supports the native Anthropic protocol and Claude Code; everything else is one OpenAI-compatible endpoint: https://byesu.si/v1.
Aisle 01 — Image (posted per-image prices)
Images are billed per image — real numbers, on the shelf, the same at 1K, 2K and 4K. Requests that fail are not charged.
| Model | What it is | Price |
|---|---|---|
| gpt-image-2 | OpenAI GPT Image 2, text→image & edits, up to 6 reference images | $0.009 / image |
| gpt-image-2.5 | OpenAI GPT Image 2.5 (newer generation), up to 6 reference images | $0.009 / image |
| gemini-3.0-pro-image-ultra | Google Gemini 3 Pro Image (Nano Banana Pro) | $0.009 / image |
| gemini-nano-banana-2.1-ultra | Google Nano Banana 2.1 | $0.008 / image |
| gemini-3.1-flash-image-ultra | Google Gemini 3.1 Flash Image (Nano Banana 2) | $0.007 / image |
The hook, spelled out: Google lists Nano Banana Pro at about $0.24 per 4K image. On byesu, gemini-3.0-pro-image-ultra is $0.009 at 4K — so 100 4K images come to about $0.90, versus roughly $24 at Google's standard list rate (Gemini API pricing, checked 2026-10-22).
gpt-image-2 is a flat $0.009 / image; calling OpenAI directly the same model is token-billed (~$0.006–$0.211 per 1024×1024 image by quality, more at 2K/4K). Full side-by-side with sources on the Image API pricing page.
Video generation is not on sale at the moment. When it returns it will be billed per second, failed tasks will be refunded automatically, and prices will be posted on Video Generation.
Start generating → · Image API docs · Generation Playground
Aisle 02 — Text & coding (pay-as-you-go, transparent)
Text and coding models are pay-as-you-go with transparent multipliers on the official upstream — no model downgrade:
- Per token, input 1× / output 5×, and cache hits as low as 0.1× — for long system prompts and long documents that read the same context repeatedly, cache hits meaningfully cut real cost.
- You pay a fraction of each provider's official list price; the console shows the actual billed price per request next to the official list price, live.
- One account covers text and image from a shared balance — image calls use a token created in the media / media-gen group, no separate account needed.
Text prices vary by group and are shown live (input / output / cache) on the Model Marketplace when you sign in. See Pricing & Billing for how the multipliers work.
How it works
- Connect your key — drop one base URL (
https://byesu.si/v1) + token into the SDK or client you already use. - Access every model — one endpoint fans out to text and image models; switch with a string.
- Pay as you go — text per token, images per image. No seats, no minimums, no monthly lock-in.
- See the real price — the console records your actual charge per request, compared live against the official list price.
Quick Start (3 steps) · Choosing a group · Media API
Why byesu vs OpenRouter and the proxies
- Text and images on one account — byesu covers Claude/GPT/Gemini/Grok plus image generation from the same shared balance (image models use a media-group token), at a flat per-image price instead of resolution-based pricing.
- Official upstream, no downgrade — official channels, not reverse-engineered or mirrored; model specs match the originals.
- Real image prices, posted — $0.007 to $0.009 per image, the same at 1K, 2K and 4K.
- Cache hits as low as 0.1× — genuine savings for long-context, heavy usage.
- Six languages, local payment — docs in six languages; Alipay / USDT.
See the full breakdown: byesu vs OpenRouter · Image API pricing · Claude Opus 5.5 vs GPT-6.1 Sol · Opus 5.5 vs Sonnet 5.5
FAQ
How much does the gpt-image-2 API cost on byesu?
$0.009 per image at 1K, 2K or 4K, pay-as-you-go, no subscription or minimums. Calling OpenAI directly the model is token-billed (~$0.006–$0.211 per 1024×1024 image by quality, per OpenAI's image generation guide, checked 2026-10-22).
Which image model is the cheapest?
gemini-3.1-flash-image-ultra (Nano Banana 2) at $0.007 per image, then gemini-nano-banana-2.1-ultra at $0.008. gpt-image-2, gpt-image-2.5 and gemini-3.0-pro-image-ultra (Nano Banana Pro) are $0.009 per image.
Is there a video generation API right now?
Not at the moment — as of 2026-10-22 no video model is on sale. When video returns it will be billed per second with automatic refunds for failed tasks, and Video Generation will list the models and prices.
How is text and coding pricing handled?
Pay-as-you-go with transparent multipliers (input 1×, output 5×, cache hits as low as 0.1×) on the official upstream models with no quality downgrade. You pay a fraction of each provider's list price; the console shows the actual billed price per request.
Do I need a subscription or minimum commitment?
No. No subscriptions, contracts or minimums — top up a balance and pay per request (text per token, images per image).
How do I set up a token for image models?
Image models are only available under the media / media-gen group. Create a token in that group, otherwise you get a no available channel error. See Choosing a group.
Get started: Get your API key · Quick Start · Pricing & Billing · Image generation
Image prices are per image and shown as of 2026-10-22; the Model Marketplace is authoritative for live prices. Official list prices cited (OpenAI, Google) are for context and belong to their respective providers.