NVIDIA NIM
An inference microservices platform introduced by NVIDIA that enables developers to deploy large models as callable API endpoints. It provides standardized interfaces compatible with the OpenAI format and a model catalog marketplace, reducing the cost and complexity of building inference infrastructure.
Timeline (last 90 days)
Moonshot AI's Kimi K3 is now available on the NVIDIA NIM API catalog; developers can call it for free through an OpenAI-compatible interface, which supports multimodality, streaming output, and reasoning intensity settings.
Using optimized inference engines such as NVIDIA NIM and TensorRT-LLM can yield several times the performance improvement on the same hardware.
FCC's Tiered Routing feature allows different Providers to be specified for OPUS, SONNET, and HAIKU respectively — for example, OPUS routed through NVIDIA NIM, SONNET using the OpenRouter free tier, and HAIKU running on a local LM Studio instance with Qwen3 Coder.
All Facts (4)
NVIDIA NIM内置了TensorRT-LLM等推理加速引擎
85%UnverifiedUsing optimized inference engines such as NVIDIA NIM and TensorRT-LLM can yield several times the performance improvement on the same hardware.
50%UnverifiedFCC's Tiered Routing feature allows different Providers to be specified for OPUS, SONNET, and HAIKU respectively — for example, OPUS routed through NVIDIA NIM, SONNET using the OpenRouter free tier, and HAIKU running on a local LM Studio instance with Qwen3 Coder.
50%UnverifiedMoonshot AI's Kimi K3 is now available on the NVIDIA NIM API catalog; developers can call it for free through an OpenAI-compatible interface, which supports multimodality, streaming output, and reasoning intensity settings.
50%