Artificial intelligence news.
Straight from the original announcements.

Hi3D Open-Sources V1.0 of Its AI 3D Model, Formerly Known as Sparc3D

October 23, 2026, 12:16By the AI News Bites Editorial TeamSource: PR Newswire

Hi3D has released Hi3D V1.0, formerly Sparc3D, as an open-source AI 3D model, making the source code and model weights available online to developers, researchers and enterprises. According to the company, it is the first production-validated open-source AI 3D model.

Key points
  • Hi3D released Hi3D V1.0, formerly Sparc3D, as open source, with source code and model weights available online.
  • The training dataset grew from approximately 500,000 to two million curated 3D assets.
  • Effective sparse-grid resolution reaches up to 1536³, and Hi3D reports a 90% slicing success rate.
  • Hi3D V3.0 offers 2048³ voxel resolution, and the company also introduced Hi3D Agent and Hi3D MCP integration.
  • Hi3D cut credit consumption by over 50%.

What Hi3D released

Hi3D announced the open-source release of Hi3D V1.0 on Oct. 23, 2026. The model was previously known as Sparc3D and, according to the company, gained recognition in 2025 for its high-resolution 3D geometry modeling. Users can get the source code and model weights online, and can also generate 3D models for free on the official Hi3D website. The company points to GitHub, a live demo and Hugging Face as access points.

Before the open-source release, the Hi3D team expanded the training dataset and tested the model in real-world production workflows, with the aim of improving its practical usability.

Key capabilities

Hi3D lists several improvements in V1.0:

  • Larger-scale training: the training dataset grew from approximately 500,000 to two million curated 3D assets.
  • Geometry reconstruction: progressive geometry reconstruction is extended to an effective sparse-grid resolution of up to 1536³.
  • Structural consistency: the model preserves fine structures and emphasizes geometric consistency across multiple viewpoints.
  • Practical usability: according to Hi3D, it achieved the highest multi-view fidelity and detail richness scores in the reported evaluation, alongside a 90% slicing success rate.

The evaluation details are not described in the announcement, so these results reflect the company's own reporting.

Intended uses and the wider Hi3D ecosystem

Hi3D says the open-source model offers a shared technical foundation for applications such as 3D printing, product and e-commerce, industrial design, VR/AR/XR, and animation and film. Enterprises and developers can explore private deployment, integrate the model into existing workflows, and build specialized products and services under the applicable license.

The company describes the release as part of a broader vision to build foundational infrastructure for AI 3D creation. It continues to develop its in-house models. Hi3D V3.0, introduced as the first commercially available AI 3D model with 2048³ voxel resolution, is said to deliver finer geometric detail, more coherent complex structures and improved texture quality.

Hi3D also introduced Hi3D Agent, a creator-first 3D agent. It combines large language model reasoning and tool use with Hi3D's proprietary 3D model, with the goal of simplifying workflows and lowering technical barriers. A companion Hi3D MCP integration makes the company's capabilities available to external agents and applications.

What it means for creators and developers

For developers and researchers, the release means direct access to the code and weights of a model the company says has been tested in production workflows. Enterprises can look at private deployment, subject to the license terms that apply. The announcement does not specify those terms, so anyone planning commercial use will need to check the license.

Alongside the release, Hi3D announced a reduction of over 50% in credit consumption, which the company says lowers the barriers to 3D creation for more creators.

Related news

Products and tools
Products and toolsNovember 12, 2026, 20:31

Snowflake Adds Inkling 2 Small to Cortex Training

Snowflake has made Inkling 2 Small available in Cortex Training public preview, giving teams a way to benchmark, post-train, evaluate and serve the open-weight model on…

Google brings offline speech recognition and TTS to AI Edge Gallery
Products and toolsNovember 10, 2026

Google brings offline speech recognition and TTS to AI Edge Gallery

Google is adding on-device Automatic Speech Recognition and Text-to-Speech to Google AI Edge Gallery, allowing developers to build voice-enabled AI experiences that work…

Pika launches open-source H3 Flex video model with real-time generation and 4K output
ModelsOctober 29, 2026

Pika launches open-source H3 Flex video model with real-time generation and 4K output

Pika has released H3 Flex, an open-source omni video model family post-trained on MiniMax H3. According to Pika, it ranks #1 among open-source text-to-video models and…