There's no information at all, this is chaos

AI & Data Science NVIDIA NIM Models
1

I’ve been developing a project using NVIDIA’s APIs for the last few days. First of all, thank you for providing them for free—it’s an amazing resource for developers.

I only have one suggestion: please improve communication.

Today GLM-5.1 was removed, but I couldn’t find any information about when GLM-5.2 will be available. DeepSeek V4 Pro also took several minutes to respond, with no indication that the model was overloaded. As a developer, it’s impossible to know whether the issue is in my code or on the API side.

A simple roadmap for model deprecations and replacements, plus a status page showing model-specific issues (high latency, overload, etc.), would save developers a lot of time.

Thanks again for making these APIs available.

2

I also immensely appreciate NVIDIA providing this for free to us. That being said, I don’t even feel I have the right to demand anything at all, other than to just say thank you. If you are a developer, you can simply develop something to notify you, or just look at the reason in your request response:

400 - Malformed JSON, invalid parameters, non-existent model
400 + DEGRADED - NVIDIA service temporarily degraded
401 - Invalid/expired API key
403 - Key does not have access to the model
429 - Rate limit exceeded
500 - NVIDIA internal error
502 - Bad gateway — upstream unavailable
503 - Service temporarily unavailable
504 - NVIDIA gateway timeout AbortError
ECONNREFUSED/DNS - Network down, NVIDIA offline, DNS failed