Frontier models behind one key
Chat with GPT 6.1 Terra, Muse Spark 1.3 Contributor, DeepSeek V4.1 Flash (1017), Gemini 3.8 Flash and GLM 5.3 FlashX through any OpenAI SDK, and generate images, video, music and speech with the same key. Every request is paid from your Hoosha credit balance.
Make your first call
Create a key
In the Hoosha developer console (
console.hoosha.com). It starts withsk-hoosha-, is shown once, and works for chat and media alike.Point the OpenAI SDK at Hoosha
Set the base URL to
https://api.hoosha.com/v1. The rest of your chat code stays as it is.Check what it cost
Every response’s
usageincludescost_credits, the exact price of that request in credits. Streams send it when you setstream_options.include_usage.
Code
Models
Chat
Credits per million tokens
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| GPT 6.1 Terragpt-6.1-terra | 922K | 121.33 | 6.07 | 606.63 |
| Muse Spark 1.3 Contributormuse-spark-1.3-contributor | 1M | 12.13 | 0.24 | 24.27 |
| DeepSeek V4.1 Flash (1017)deepseek-v4.1-flash-1017 | 1M | 18.20 | 0.36 | 72.80 |
| Gemini 3.8 Flashgemini-3.8-flash | 1M | 90.99 | 9.10 | 454.97 |
| GLM 5.3 FlashXglm-5.3-flashx | 1M | 44.89 | 10.92 | 151.66 |
Decisions
Typed answers with probabilities. Credits per million input tokens; output is free.
| Model | Context | Input | Output |
|---|---|---|---|
| Jev 1.14jev-1.14 | 262K | 5.10 | 0.00 |
Media
Charged per job
- Image
- Nano Banana 2.1, Nano Banana 2, ChatGPT Image 2.5 Flare, ChatGPT Image 2.5 Sunburst, Grok Imagine 2, ChatGPT Image 2.0, Hoosha V1, ChatGPT Image 1.5, Flux Kontext and SeeDream 4.5
- Video
- Grok Imagine 1.5, Kling 3.0 Turbo, Seedance 2.5, Seedance 2, Kling Motion, Genjutsu, Google Omni, Veo 3.1, Runway, Kling 2.5 Turbo, Seedance v1 and Lipsync
- Music
- Suno V6
- Speech
- ElevenLabs v4, ElevenLabs v4 Turbo and ElevenLabs v3
How credits work
Chat
A request holds its largest possible cost from your balance, then settles to the exact cost of the tokens used when the model finishes. Fractions of a credit carry over to your next request, and a request that fails costs nothing.
Media
A job is charged when it is accepted, and the response shows the price. If the provider fails the job, the credits return to your balance automatically, and per-second video models refund the seconds they did not deliver.