4B Model Arko-T Beats GPT-5 in Text-to-CAD Generation
机器之心 · wechat · 2026-07-31
Bit Infinite released Arko-T, a 4B parameter foundation model for structured 3D generation. Unlike traditional visual 3D generation, Arko-T outputs executable Build123d programs to generate CAD entities with named parameters, structural features, and modeling history, solving the limitation of generated models being unusable for engineering modifications.
Performance-wise, Arko-T outperformed seven top-tier LLMs including GPT-5.2 and Claude-4.5-sonnet, taking first place in 8 out of 12 quantitative metrics. It also offers extremely fast inference and low costs: averaging 0.41 seconds per sample locally, with a full benchmark cost of just $0.28, significantly cheaper than API models.
Beyond the model, the team's long-term goal is to build a physical 3D data engine. By connecting with 3D printing users and the Maker community, they aim to collect real-world design modification feedback to continuously improve the model. Ultimately, this structured data will be converted into OpenUSD assets for robot simulations, providing crucial physical data infrastructure for embodied AI and world models.
More from Multimodal
- Mind-Blowing AI Concept: Visualizing Pigeon 'Flight-Mile' Delivery to Dodge Traffic — sivakumaranvlsi · 2026-07-31
- MiniMax H3 Video Model Now Available on Vercel AI Gateway — evilrabbit_ · 2026-07-31
- Testing Flux 3: Generating Synchronized Split-Screen Videos via Complex Prompts — umesh_ai · 2026-07-31
- MPIE-Bench: Evaluating Anatomical Errors in Multi-Person Image Editing — muset-ai · 2026-07-31
- RefCaptioner: Grounding Video Captions to Multiple Reference Images — KlingTeam · 2026-07-31
- MiniMax H3 Tops Video Editing Leaderboard, Undercutting Rivals at $7.80/min — ArtificialAnlys · 2026-07-31