Hello NVIDIA team,
I’m building a personal project that uses the NIM hosted API for both text and image generation. Text models work fine, but every image generation model I’ve tried returns a 404 error tied to my account, and the deprecation notice on the model page doesn’t point to a working replacement. Details below.
Account info:
- Account ID:
68X6Mfwnqn5FfpLzIs-fjNlCY0q5UMuotOGWalpiocU - API key prefix:
nvapi-7ehBy...(key works fine for chat/completions models)
What works:
curl -X POST “nvidia.com”
-H “Authorization: Bearer $NVIDIA_API_KEY”
-H “Content-Type: application/json”
-d ‘{“model”: “meta/llama-3.1-8b-instruct”, “messages”: [{“role”: “user”, “content”: “hi”}], “max_tokens”: 50}’
This returns a normal successful completion. **What fails — Stability AI models (sdxl-turbo, stable-diffusion-xl, stable-diffusion-3-medium):**
curl -X POST “nvidia.com”
-H “Authorization: Bearer $NVIDIA_API_KEY”
-H “Content-Type: application/json”
-d ‘{“text_prompts”: [{“text”: “a cat sitting on a chair”, “weight”: 1}], “seed”: 0, “sampler”: “K_EULER_ANCESTRAL”, “steps”: 2}’
response:
{“status”:404,“title”:“Not Found”,“detail”:“Function ‘f886140c-424e-4c82-a841-99e23f9ae35d’: Not found for account ‘68X6Mfwnqn5FfpLzIs-fjNlCY0q5UMuotOGWalpiocU’”}
The model page itself for sdxl-turbo now shows a deprecation notice:
“This NIM Endpoint has been deprecated. Please transition to another model to avoid any service interruptions.” — but it doesn’t link to which model I should switch to for the hosted free-tier API specifically. Same 404 “Function not found” pattern occurs for `stabilityai/stable-diffusion-xl` and `stabilityai/stable-diffusion-3-medium` with the same account.
**What fails — Qwen-Image:**
curl -X POST “nvidia.com”
-H “Authorization: Bearer $NVIDIA_API_KEY”
-H “Content-Type: application/json”
-d ‘{“model”: “qwen/qwen-image-2512”, “prompt”: “a cat sitting on a chair”, “n”: 1, “response_format”: “b64_json”}’
Same 404 “Function not found for account” error.
I also tried the same Qwen-Image request against the `://nvidia.com` subdomain and got a plain “404 page not found” (not even a JSON error), suggesting that route doesn’t exist on that subdomain at all.
**Questions:**
1. Is hosted (non-self-hosted, non-Docker-container) image generation still available at all on the free/personal developer tier in June 2026? The Visual GenAI NIM docs only show localhost:8000 examples, which look like they’re for self-hosted containers requiring 48–80GB VRAM GPUs, not the hosted API catalog.
2. If hosted image generation is still available, which model ID and which exact base URL should I be using for Qwen-Image or any other current image model on the free tier?
3. Is there a way to query which models/functions are actually enabled for my specific account, similar to what was asked in related threads regarding listing supported models via API? Thanks in advance — happy to provide any additional logs or test calls if useful.