Apex Neural Rendering arrived in beta

Apex Compute announced a cross-platform suite called Apex Neural Rendering (ANR) on November 11, 2026. According to the company's statement on X, the package brings together Super Resolution, Ray Reconstruction and Frame Generation; The part currently available for download is the frame generation (ANR FG) beta. Super Resolution and Ray Reconstruction are still in training. Apex Compute claims up to 2.4x faster performance and higher quality compared to NVIDIA DLSS; These figures belong to the company's own measurement.

What is Apex Neural Rendering and how does it work?

According to the official Apex Neural Rendering page, the technology can run on NVIDIA GeForce RTX 20 series and newer GTX 16, AMD Radeon RX 5000 series and newer Steam Deck, Intel Arc A and B and Apple silicon M1 and later systems. Supported environments are Windows 10 and 11, Linux and SteamOS (with Proton), and macOS 26 or later.

There are two packages for the beta: anr-fg-full, for games that don't use OptiScaler; anr-fg-addon for games that already use OptiScaler. According to the installation note, frame generation in the game is turned on via the NVIDIA option offered by the game. This doesn't mean it's official integration from the game publisher; A beta installed via OptiScaler.

ANR FG packages and release notes are published on the company's GitHub repository. FG V.B13, released on November 12, 2026, also adds an Apple silicon test kit for macOS 26 or later. This kit is offered to measure ANR FG on Metal in the same frames as Apple MetalFX frame interpolation.

What is the 2.4 times claim based on?

In its table on the official site, Apex Compute writes 1.01 ms for ANR FG and 1.26 ms for NVIDIA DLSS Frame Generation per frame produced with NVIDIA GeForce RTX 5090 and 3840 x 2160 resolution. In the same table, the average time in 4X mode is 0.61 ms in ANR FG and 0.82 ms in DLSS FG. This difference is around 1.25 times; It should not be lumped together as the same metric as the 2.4x claim on X.

In the separate graph shared by the company on the The unit is not written on the chart. The 2X pair (0.34 to 0.81) remains close to the company's 2.4x cap; However, this is the company's own chart, not an independent test. For NVIDIA's own frame generation evolution, check out our DLSS 4.5 announcement.

What does the quality and memory table show?

The quality table on the official site also compares ANR FG with DLSS FG. In the 2X line, the CGVQM-5 mean patch value is 96.59 on the ANR side and 93.66 on the DLSS side; ColorVideoVDP is listed as 9.24 by 8.77 JOD, while PSNR is listed as 33.84 by 31.22 dB. The company also published 95% confidence intervals. These are not independent laboratory results.

On the same site, GPU memory at 4K is given as 214 MB for ANR FG and 562 MB for DLSS FG. The weight file is 0.17 MB. For a discussion of frame generation on different brands of cards, you can see our article on the operation of DLSS technologies on AMD graphics cards.