Models

List the models available to your account at runtime:

curl https://api.crustoff.app/v1/models \
  -H "Authorization: Bearer $CRUSTOFF_API_KEY"

Each entry includes a non-standard modality field (text or image) so you can tell which endpoint to use. OpenAI SDKs ignore the extra field.

{
  "object": "list",
  "data": [
    { "id": "qwen2.5-7b-instruct", "object": "model", "owned_by": "crustoff", "modality": "text" },
    { "id": "sdxl", "object": "model", "owned_by": "crustoff", "modality": "image" }
  ]
}

Text models

Call these with /v1/chat/completions, /v1/completions, or /v1/embeddings.

Model	Context	Notes
`glm-5.2`	1M	Z-AI flagship — long-horizon tasks
`deepseek-v4-pro`	1M	MoE reasoning & coding (1.6T total / 49B active)
`deepseek-v4-flash`	1M	Efficiency-focused MoE (284B total / 13B active)
`kimi-k2.7-code`	256K	Coding & agentic (Moonshot)
`qwen3.6-35b-a3b`	256K	Alibaba MoE (35B total / 3B active)
`qwen2.5-7b-instruct`	32K	Strong general-purpose chat model

The list is always live — check /v1/models for what your account can call right now.

Image models

Call these with /v1/images/generations.

Model	License	Best for
`sdxl`	OpenRAIL++	Fast, versatile, huge LoRA ecosystem
`flux-schnell`	Apache-2.0	Modern 12B quality in 1–4 steps
`qwen-image`	Apache-2.0	Flagship 20B quality, strong text rendering

All image models we host are commercially licensable for resale. We don’t host FLUX.1 dev — its license forbids running it as a paid service — so you won’t see it in the list.

Video models

Call these with /v1/videos/generations (async, image-to-video).

Model	License	Best for
`wan2.2-i2v`	Apache-2.0	Image-to-video — animate a still into a short clip

See Pricing & billing for per-model rates.

Authentication Text generation