OTOY has posted the first public beta of OctaneRender 2027.1, which OTOY says makes gaussian splat rendering faster on every platform.
Octane has path traced splats since OctaneRender 2026 shipped in November 2025, and the 2026.4 update in July added SPZ v4 reading and lower splat GPU memory use. OTOY says its lab tests show Beta 1 raising splat render throughput by 1.2x to 2.4x on Windows and Linux, and by 1.9x to 3.8x on macOS, compared with the previous approach. OTOY does not name the hardware or scenes behind those figures.
The speedup carries two costs. Renders show slightly more noise at equal sample counts, although OTOY reports that perceptual error, measured with FLIP, dropped in every case within the same time budget. The change also removes the 2026.1 compatibility mode, which replicated a bug affecting slightly transparent splats, so scenes that relied on that behavior may render differently.
OctaneRender 2027.1 Beta 1 also moves the real-time neural rendering mode to DLSS 4.5 and lets it use the new visualization cache. That mode does not render splats yet. OTOY says splat support for it is being worked on and planned for the next build. Render to Gaussian Splat, which the Octane 2027 roadmap described as outputting an entire scene as a splat file, does not appear in this beta.
Octane Standalone also gains a built-in MCP server for AI agents that accepts local connections only and has no authentication. The beta adds separate denoiser and upsampler nodes for OIDN, Octane AI, DLSS, FSR, MetalFX, Legacy AI, and Linear, along with USD DistantLight import.
OTOY calls the build almost feature complete and advises against using it for production work. Beta 1 requires NVIDIA driver 580 or later and supports macOS 15 through macOS 27.
The beta is available to download from the OTOY forums.
