Box3D: A New Open-Source Physics Engine from the Creator of Box2D

Erin Catto has launched Box3D, an open-source 3D physics engine based on the optimized architecture of Box2D. Developed to solve stability and performance issues in 'The Legend of California,' the engine features SIMD acceleration and advanced collision types. Although currently in alpha, Box3D is already being used in several professional game projects and is available on GitHub.

Key Points

• Box3D is a new open-source 3D physics engine built with a C17 API and a core architecture nearly identical to Box2D v3.0.
• The engine was developed to overcome limitations in Unreal's Chaos physics, specifically regarding continuous collision stability and gyroscopic torque simulation.
• It features high-performance capabilities including wide SIMD contact solvers, multi-threading hooks, and support for large worlds using doubles for position.
• The project is already integrated into several professional titles and engines, including s&box and the Esoterica animation system.
• Catto open-sourced the engine to preserve his professional knowledge across different roles and to provide a customizable tool for the game development community.

Sentiment

The overall sentiment toward Box3D is strongly positive. The community largely agrees with the article's framing that this is an important and promising open-source release, with most criticism focused on adoption questions, comparisons with existing engines, and the broader economics of open-source maintenance rather than on Box3D itself.

In Agreement

• Box3D is a welcome addition to the small set of serious open-source physics engines, especially because it comes from the creator of Box2D and inherits a reputation for readable, practical code.
• The C API, small build footprint, and similarity to modern Box2D make the engine attractive for language bindings, custom engines, and WebAssembly builds.
• Determinism, replay support, and potential rollback-friendly behavior are seen as valuable for networked games and simulations where mainstream engine physics can be difficult to control.
• Existing use by projects such as s&box and endorsement from experienced physics and game-networking developers increase confidence that Box3D is more than a toy release.
• The release broadens the open-source physics landscape and gives developers another option alongside Jolt, Bullet, PhysX, Rapier, Avian, Newton, and related web ports.

Opposed

• Some commenters want concrete comparisons with Jolt, Bullet, PhysX, Rapier, and other engines before treating Box3D as the best choice for production work.
• Several game developers argue that general-purpose physics engines can be awkward for gameplay-heavy projects where custom feel, platformer behavior, deformable objects, or soft-body effects matter more than realism.
• A few people use Box2D's history to argue that permissive open source can leave maintainers undercompensated, while others counter that license terms should define expectations and that payment obligations should be written explicitly.
• Some caution that engine-level determinism and replay are hard in practice, especially when a larger game engine or scripting layer can introduce hidden state differences.
• There is mild skepticism that Box2D aged unevenly for certain use cases, which leads some commenters to wait for higher-level abstractions or project-specific proof before adopting Box3D.