NVIDIA NIM and ChatRTX (deprecated)
NVIDIA's inference microservices; the ChatRTX demo app is deprecated
Overview
NVIDIA NIM packages open and NVIDIA models as optimised inference containers for NVIDIA GPUs, free for Developer Program members to prototype and self-host on up to 16 GPUs, and licensed through NVIDIA AI Enterprise for production. ChatRTX, its local RTX chat demo, was deprecated in January 2026.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- MIT-licensed source on GitHub
- No longer maintained
- Hosted endpoints for prototyping
- Self-host on up to 2 nodes or 16 GPUs
- Production use of NIM
- Business Standard Support included
- Production use on cloud GPUs
What you can produce with NVIDIA NIM and ChatRTX (deprecated)
- NIM inference microservices for open and NVIDIA models
- Optimised inference containers with standard APIs
- Hosted NIM endpoints on build.nvidia.com
- Self-hosted NIM on NVIDIA GPUs
- ChatRTX local document chat source code (archived, MIT)
ChatRTX is finished — NVIDIA deprecated and archived it in January 2026 — so anyone wanting private document chat on an RTX PC should use a maintained open-source alternative. NIM is the part worth knowing: a clean way to run optimised models on NVIDIA GPUs, free for development on up to 16 GPUs, with production licensed through NVIDIA AI Enterprise from $4,500 per GPU a year or $1 per GPU-hour in the cloud. That makes sense for organisations standardising on NVIDIA infrastructure; hobbyists and small teams will find the production licence steep.
What we found
This entry covers two NVIDIA ways of running language models on your own hardware. ChatRTX was a free Windows demo app that let owners of GeForce RTX graphics cards chat with their own documents, notes and photos locally, using retrieval-augmented generation built on TensorRT-LLM, LlamaIndex and FAISS, with models such as Llama 3.1 8B and voice input through Whisper and Parakeet. It needed Windows 11, a recent driver and an RTX 30 or 40 series card, an RTX 5080 or 5090, or an RTX 6000 Ada with at least 8 GB of video memory. NVIDIA deprecated it on 21 January 2026: the GitHub repository, under the MIT licence, is archived, its forum is locked and its product page redirects to NVIDIA's general AI-on-RTX hub. NVIDIA has not named a successor.
NIM, NVIDIA Inference Microservices, is the part that continues. It packages open and NVIDIA models as containers with optimised inference engines and standard APIs, to run on NVIDIA GPUs in a data centre, in the cloud or on workstations. Members of the free NVIDIA Developer Program can prototype against hosted NIM endpoints on build.nvidia.com and self-host NIM on up to two nodes or 16 GPUs for development, testing and research. Developers report a limit of about 40 requests a minute on the hosted endpoints, though NVIDIA does not publish one.
Production use, meaning serving real end users, needs an NVIDIA AI Enterprise licence: $4,500 per GPU for one year, $13,500 for three years, $18,000 for five years or $22,500 perpetual with five years of support, or $1 per GPU-hour plus instance costs through cloud marketplaces. Education and Inception start-up pricing is $1,125 per GPU a year, and a five-year subscription comes with H100, H200 NVL and A800 PCIe GPUs.
NVIDIA, founded in 1993 and based in Santa Clara, California, sells the hardware all of this runs on. For local document chat, look to maintained open-source tools; for production inference on NVIDIA GPUs, NIM is the supported route.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including NVIDIA NIM and ChatRTX (deprecated)'s official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with NVIDIA NIM and ChatRTX (deprecated) unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to NVIDIA NIM and ChatRTX (deprecated) directly →
Spotted something out of date? Suggest an update →
Alternatives to NVIDIA NIM and ChatRTX (deprecated)
More in Code & Dev Tools
