Model Deprecation Request

AI & Data Science NVIDIA NIM Models
, , , ,
1

Dear NVIDIA NIM Team,

Thank you for maintaining and continuously improving the NVIDIA NIM API platform. I would like to submit a request regarding several models that appear to be outdated or superseded by newer alternatives.

As the AI ecosystem evolves rapidly, updating the available model catalog would improve user experience, performance, and overall platform competitiveness. I would like to suggest considering the deprecation or replacement of the following models:

  • Kimi 2.6 → Upgrade to Kimi K2.7 or Kimi K3.

  • MiniMax M2.7 → Replace with an optimized MiniMax M3 deployment.

  • NVIDIA Nemotron Nano 12B V2 VL (nvidia/nemotron-nano-12b-v2-vl).

  • Qwen3 Next 80B A3B Instruct (qwen/qwen3-next-80b-a3b-instruct) → Consider upgrading to Qwen 3.6 36B A3B or other models from the Qwen 3.6 family.

  • Llama 3.1 Nemotron Nano VL 8B V1 (llama-3.1-nemotron-nano-vl-8b-v1).

  • Meta Llama 3.2 90B Vision Instruct (meta/llama-3.2-90b-vision-instruct).

  • BGE-M3.

Suggested replacement priorities:

  1. Kimi K3 (or K2.7) for stronger reasoning and coding capabilities.

  2. MiniMax M3 for improved efficiency and multimodal performance.

  3. Qwen 3.6 family models (particularly 36B A3B) for a modern balance of quality, latency, and cost.

  4. Newer Nemotron and Llama variants where applicable.

  5. Updated embedding models to replace BGE-M3 if more capable alternatives are available.

Benefits of these upgrades include:

  • Better reasoning and coding performance.

  • Improved multimodal capabilities.

  • Lower latency and higher throughput.

  • Reduced maintenance overhead for legacy deployments.

  • Greater alignment with current open-model benchmarks and community adoption.

The NVIDIA NIM platform is already an excellent ecosystem for deploying AI models, and keeping the model catalog aligned with the latest generation of foundation models would further strengthen its position for developers and enterprises alike.

Thank you for considering this feedback. I appreciate the team’s efforts and look forward to future updates to the NVIDIA NIM API platform.

Best regards,

supervisorxscp

2

Hi there,

The build team is constantly trying to bring the latest updates.

Stay tuned!

Best,

Aharpster

3

Thanks all the team :D

4

Hello Aharpster. If it is not too much trouble. Could you please tell us why both Deepseek V4 Pro & Flash where deprecated today please?.

Many people here used them, including myself…

5

Probably to free up servers for the Kimi K3, as well as the updated Deepseek V4 Flash. It would be nice if they gave a seven-day notice before removing models, but I understand why they’re doing it this way. They can’t keep 6-10 big models while adding 2-3 other big models. Deepseek v4 Flash will be back in 1-3 days, and they might also update Kimi to K3 this week.

I also understand why the Nvidia NIM team does not explain anything or does not respond to anyone, due to the disgusting attitude towards them from abusers and free users.

6

Damn, this is why we cant have free stuff