
Rerun
Rerun is an open source SDK and viewer for multimodal, time-series robotics data, and version 0.36.0 added an experimental Gaussian splat archetype with its own EWA splatting renderer.
Location
Stockholm, Se
Size
Total raised
Platform
Desktop viewer on Windows, macOS and Linux plus a browser based web viewer, with Python, Rust and C++ SDKs
GPU
No minimum GPU is published. The viewer renders through wgpu and defaults to Vulkan on Windows and Linux and Metal on macOS, while the web viewer prefers WebGPU and falls back automatically to WebGL. Rerun documents known driver issues and ships a renderer_gl flag for machines where Vulkan is unavailable
Pricing
The SDK and viewer are free and open source. Rerun Hub, the managed backend, is commercial with no published price and routes through Contact us
License
Dual licensed Apache 2.0 and MIT
Best for
Inspecting multimodal robotics and reconstruction data on one shared timeline
Updated
July 2026
Rerun is an open source SDK and viewer for multimodal, time-series data, built by Rerun Technologies AB, a remote-first company whose founders are based in Stockholm. It describes itself as the data layer for physical AI, covering logging, storage, query, viewing and training on multi-rate multimodal data in one toolchain. The SDK is dual licensed Apache 2.0 and MIT, carries Python, Rust and C++ APIs alongside a browser based web viewer, and its core is written in Rust. The repository stands at 11.3 thousand stars.
Gaussian splatting arrived in version 0.36.0 as an experimental GaussianSplats3D archetype with a PLY loader, drawn by a new EWA splatting renderer in re_renderer with spherical harmonics to degree 3. A gaussian splat PLY file opens directly in the viewer, and the archetype can equally be logged from any of the three SDKs, taking centers as Position3D with scales, quaternions and colors recommended and sh_coefficients optional. Unlike a training checkpoint it stores values in their natural form, linear scales and base colors with opacity as RGBA. Rerun marks the archetype unstable and warns that the data may not stay backwards compatible, and PLY is the only splat format the release names.
The recording model is why a splat archetype belongs here at all. Rerun stores multi-rate multimodal data as column chunks in rrd files, so a reconstruction can sit on the same timeline as the camera frames, poses and depth that produced it, and the whole recording can then be queried with SQL or a dataframe API rather than exported. That is the same consumption pattern robotics simulators have converged on, where Isaac Sim 6.0 reads NuRec scenes directly.
The viewer installs through pip, cargo, conda, pixi or a community maintained snap, and rerun-cli binaries are published as GitHub release artifacts. Rerun Hub is the commercial layer, a managed catalog over the customer's own S3 compatible storage adding byte range indexing, dataset mixes streamed to GPUs, link sharing, auth and SSO, and single tenant isolation in a chosen region. It is sold through Contact us and carries no published price.
Frequently Asked Questions
Is Rerun free?
The Rerun SDK and viewer are free and open source, dual licensed under Apache 2.0 and MIT. Rerun Hub, the managed production backend, is a separate commercial product with no published price that routes through Contact us.
Does Rerun support Gaussian splatting?
Yes, as of version 0.36.0, through an experimental GaussianSplats3D archetype with a PLY loader and a new EWA splatting renderer that handles spherical harmonics to degree 3. Rerun marks the archetype unstable and warns the data format may not remain backwards compatible, and PLY is the only splat format the release names.
What GPU does Rerun need?
Rerun publishes no minimum GPU. The viewer renders through wgpu and defaults to Vulkan on Windows and Linux and to Metal on macOS, while the web viewer prefers WebGPU and falls back automatically to WebGL where WebGPU is unavailable. Rerun documents known driver issues and provides a renderer_gl flag for machines that cannot run Vulkan.
What languages does Rerun support?
Rerun ships SDKs for Python, Rust and C++, and its core is written in Rust. The Python package installs with pip install rerun-sdk, and the viewer can also be installed on its own through cargo, conda, pixi or a community maintained snap.
What is Rerun Hub?
Rerun Hub is the commercial backend for the Rerun data layer. It provides a persistent managed catalog over a customer's own S3 compatible object storage, with byte range indexing, SQL and dataframe queries into recordings, derived columns without copies, dataset mixes streamed directly to GPUs for training, and single tenant isolation in a chosen region.
Alternatives
Recent Headlines
Scaniverse #SplatYourWorld Challenge Winners Announced
Rerun Adds an Experimental Gaussian Splat Archetype and EWA Renderer in 0.36.0
XGRIDS Adds a Gaussian Splat Format Converter to LCC Studio V2.3.0
Insta360 Brings Gaussian Splatting to the X6 With Spatial Capture
SideFX Labs Ships Three Gaussian Splat Nodes for Relighting, Delighting, and Normal Reconstruction
Esri's Arkadiusz Szadkowski on Splats at National Scale, Formats, and Finding the Early Adopters
Splatlight Gives Gaussian Splats Self-Shadowing Under Scene Lights in Blender
NouerA's ".exe" Uses 4D Gaussian Splatting — With 21 Cameras in Seoul
Captures Studio Releases Multi Splat for Aligning, Comparing and Viewing 3DGS
Esri Adds Georeferenced Gaussian Splat Layers to the ArcGIS Maps SDK for Unreal Engine at 2.4

