The hosted TRELLIS endpoint appears to be down. Every request returns
HTTP 500 “Internal Server Error” with response header nvcf-status: errored.
This includes the exact Shell snippet published on the model page
(https://build.nvidia.com/microsoft/trellis?snippet_tab=Shell), using the
built-in example image with no modifications:
POST https://ai.api.nvidia.com/v1/genai/microsoft/trellis
{
“image”: “data:image/png;example_id,0”,
“slat_cfg_scale”: 3,
“ss_cfg_scale”: 7.5,
“slat_sampling_steps”: 25,
“ss_sampling_steps”: 25,
“seed”: 0
}
Text-mode requests (prompt only) and uploaded NVCF assets fail identically.
Authentication is fine — the same API key works for FLUX.1-dev,
FLUX.1-Kontext-dev, and Nemotron on the same account, and credits are
being consumed normally by those models.
Sample failing request ID: nvcf-reqid b649d4e2-fc19-4e5b-b6f5-51a0bb38ff1b
Timestamp: 2026-07-11 07:29 UTC
Could you check the health of the hosted TRELLIS function?