VGGT-Omega (fastforks)

VGGT-Omega (Meta), the successor of VGGT: cameras and depth maps of every view from unposed images in one forward pass. A DINOv3 ViT-L/16 patch embedding feeds 24 frame and 24 inter-frame (global or register-only) attention blocks with q/k norm and 2-D RoPE; a camera head and a DPT dense head decode the tokens. The checkpoint is gated on Hugging Face (accept the FAIR licence there, then put the file on FASTFORKS_WEIGHTS_PATH); there is no mirror.

This repo is the model card only: it holds no weights. fastforks downloads the authors' weights from the links below and checks them against the SHA-256 in the model card. Code: https://github.com/kleinicke/fastforks. Kernels for the kernels backend: https://huggingface.co/fastforks/kernels. Original code: https://github.com/facebookresearch/vggt-omega at 48b23c8.

Author weight repos: https://huggingface.co/facebook/VGGT-Omega.

Task: multiview. Tier: experimental. Outputs: depth, confidence, intrinsics, cam_to_world. Backends: portable, kernels.

Input: several images of one scene (posed and calibrated for multi-view stereo, unposed otherwise). Output: depth maps and, where the model predicts them, cameras and 3-D points.

Variants

Weights are the authors' files (not hosted here). SHA-256 shortened to 12 hex digits; the full hashes are in card.toml.

variant	weights (author link)	SHA-256	weights licence	commercial use
vggt-omega-1b-512@1	vggt_omega_1b_512.pt	c02da418b18b	LicenseRef-FAIR-Noncommercial-Research	no

Licence

Original-code strings, recorded as they are (research-use package; check them before any other use). Code: LicenseRef-FAIR-Noncommercial-Research. Files: LICENSE.

Training data: Per the repository README: a mixture of public multi-view datasets (uCO3D annotations released); DINOv3 initialisation. The 512 checkpoint is recommended for in-the-wild use, not benchmarking (vggt_omega_1b_416_reproduce.pt is the benchmark reference)..

• Licence note: Code and weights are under the FAIR Noncommercial Research License; the checkpoint is gated on Hugging Face (accept the licence there) and not mirrored.

Licence evidence: https://github.com/facebookresearch/vggt-omega/blob/48b23c8/LICENSE and https://huggingface.co/facebook/VGGT-Omega/blob/1041e80fc0e911235d3426b0a3d9a81075111a53/LICENSE.txt

Input size

any H x W >= 1, resized or padded inside the model Tested up to (portable, fp16, RTX 5060 Ti 16 GB): vggt-omega-1b-512 3000x4000 (5.5 GiB peak allocated, 6.7 reserved)

Speed and memory

Multiview, synthetic, 336x518

model	time (ms): original fp32	original	portable	kernels	kernels speed-up: vs original	vs portable	peak memory (MiB, allocated / reserved): original	portable	kernels
vggt-omega-1b-512	2650	906	726	464	2.0x	1.6x	7495 / 7654	3421 / 3518	3422 / 3518

Startup and first call

model	load (s): original	portable	kernels	first call (s): original	portable	kernels
vggt-omega-1b-512	16.31 (336x518)	3.31 (336x518)	2.27† (336x518)	1.10 (336x518)	0.73 (336x518)	0.46 (336x518)

torch.compile and TensorRT

model	input	portable ms	compile ms	compile s (cold)	TensorRT fp16 ms	ONNX + build s	kernels ms	peak MiB (alloc / reserved): portable	compile	TensorRT (device)
vggt-omega-1b-512	336x518	726	588.6	65.7	38.5	174	464	3421 / 3518	3082 / 3154	2664

Fixed-size speed and peak memory, all models

model	mode	480x640	720x1280	1080x1920	12 MP 3000x4000
vggt-omega-1b-512	portable	722.8 ms / 3449 / 3866	723.2 ms / 3559 / 4038	725.7 ms / 3771 / 4366	757.0 ms / 5594 / 6894
vggt-omega-1b-512	kernels	461.8 ms / 3449 / 3796	461.1 ms / 3560 / 3968	463.9 ms / 3771 / 4332	493.0 ms / 5594 / 6894

Fixed-size cells: time (ms) / peak allocated / peak reserved memory (MiB).

Hardware, protocol and caveats: benchmarks (RTX 5060 Ti 16 GB, sm_120). Original = the authors' code at the precision of portable (fp16, or (fp32) where fp16 is unsupported or not finite); original fp32 = its default fp32 run. Peak memory is the torch allocator's.

Accuracy

Parity with the original code

Every variant is checked against reference outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.02 and p99 <= 0.25 (absolute error of the output depth).

Usage

from fastforks import multiview

# non-commercial weights: run with FASTFORKS_NONCOMMERCIAL=1 in the environment
model = multiview.load("vggt-omega-1b-512")  # mode="auto": the preferred backend from recorded measurements
result = model(["view0.jpg", "view1.jpg", "view2.jpg"])  # the views of one scene, no cameras needed
result.depth  # [V, H, W] float32 tensor; result.kind says what the maps mean
result.backend  # what ran, e.g. Backend(mode='portable', device='cuda:0', dtype='float16')

Install with pip install fastforks (see https://github.com/kleinicke/fastforks). Other variants: vggt-omega-1b-512.

Kernels backend

multiview.load("vggt-omega-1b-512", mode="kernels") runs our CUDA kernels (fastbb, vggt_omega). It needs a prebuilt kernel for your GPU architecture (downloaded from https://huggingface.co/fastforks/kernels on first use) and a recent NVIDIA driver; if there is none it raises BackendUnavailableError with the reason and never falls back silently. mode="auto" picks kernels when it can run and logs the choice.