# GPT-6 Astra 3D Generator

> GPT-6 Astra is OpenAI's general frontier model, rolled out from 3 September 2026; it builds 3D by planning and writing the geometry rather than being a dedicated 3D model. Kenerate AI's 3D Studio runs it as a 3D job: a brief of up to 4,000 characters or up to 10 photos returns one GLB with PBR textures.

URL: https://kenerateai.com/model/gpt-6-astra
Publisher: Kenerate AI (https://kenerateai.com)
Last updated: 2026-10-01

## Key facts
- What it is: OpenAI's general model, run here as a 3D job (not a dedicated 3D model)
- Released: Rolling out from 3 Sep 2026
- Inputs: Prompt up to 4,000 characters · 1–10 photos (+ optional prompt)
- Output: One GLB with PBR textures; can be a small scene
- Size: Longest side set in metres (10 cm to 20 m in the studio)
- Speed: Slowest engine — our runs: 3.5–8.6 min; long jobs can take more
- Engine Meshy 7.1 (Meshy): inputs Text/Photo/Multi-view; files GLB, FBX, OBJ, USDZ; typical about 3–4 min; Production assets with the finest Meshy geometry (2K / 4K), smart topology, rigging and animation.
- Engine Meshy 7 (Meshy): inputs Text/Photo/Multi-view; files GLB, FBX, OBJ, USDZ; typical about 3–4 min; Detailed game and product assets, with an Ultra detail switch, low-poly mode and auto-rigging.
- Engine Meshy 6 (Meshy): inputs Text/Photo; files GLB, FBX, OBJ, USDZ; typical about 3–4 min; Reliable textured meshes with a polygon budget you control — the engine 3D Studio started on.
- Engine Tripo P2 (Tripo): inputs Text/Photo; files GLB, FBX; typical about 3–4 min; Crisp, game-ready meshes with pro texture tiers up to Extreme and clean quad remeshing.
- Engine Tripo H3.1 (Tripo): inputs Text/Photo/Multi-view; files GLB; typical about 4–5 min; High-quality meshes, with detailed geometry and a four-view mode for exact shapes.
- Engine GPT-6 Astra (OpenAI): inputs Text/Photo/Multi-view; files GLB; typical about 5–6 min; OpenAI's general GPT-6 Astra model run as a 3D job — follows long, detailed briefs and up to 10 reference photos, returns a GLB with PBR textures at real-world size. The slowest engine.
- Engine Rodin 2.5 (Hyper3D): inputs Text/Photo/Multi-view; files GLB, FBX, OBJ, USDZ, STL; typical about 2–3 min; Studio-grade meshes from 2K to 500K faces, quad or triangle, with a HighPack for 4K textures.
- Engine Rodin 2 (Hyper3D): inputs Text/Photo/Multi-view; files GLB, FBX, OBJ, USDZ, STL; typical about 2–3 min; Clean, detailed meshes — the first photo drives the materials.
- Engine Hunyuan 3D 3.1 Rapid (Tencent): inputs Text/Photo; files GLB; typical about 1 min; The fastest geometry — an untextured (clay) GLB for quick blockouts, drafts and ideas.

## Steps
1. Open 3D Studio — Pick Text to 3D, Image to 3D or Multi-view.
2. Pick GPT-6 Astra — Choose GPT-6 Astra in the engine list.
3. Write the brief — List parts, materials, colours and proportions; add up to 10 photos.
4. Set size and wait — Choose the real-world size, generate, and check back in a few minutes.

## In depth

### A general model doing 3D
OpenAI rolled out GPT-6 Astra on 3 September 2026 as its frontier model for coding, computer use and professional work [31]; its published model card lists text output rather than a 3D modality [32]. Press coverage describes it reconstructing objects by writing CAD code and operating tools such as Blender [33].

### How the studio runs it
WaveSpeed wraps Astra in text-to-3D and image-to-3D endpoints that return one GLB with PBR textures, with the size set in metres; WaveSpeed notes that exact likeness and physical accuracy aren't guaranteed [34].

### When to choose it
Choose Astra when the brief is long and precise or you have many reference photos. For speed, drafts or specific file formats, Hunyuan, Tripo or Rodin fit better.

References:
[1] Wikipedia — STL (file format): https://en.wikipedia.org/wiki/STL_(file_format)
[2] Wikipedia — Wavefront .obj file: https://en.wikipedia.org/wiki/Wavefront_.obj_file
[3] Wikipedia — FBX: https://en.wikipedia.org/wiki/FBX
[4] Khronos Group — glTF: https://www.khronos.org/gltf/
[5] 3MF Consortium — 3MF specification: https://3mf.io/specification/
[6] Wikipedia — Universal Scene Description (USD / USDZ): https://en.wikipedia.org/wiki/Universal_Scene_Description
[7] Wikipedia — PLY (file format): https://en.wikipedia.org/wiki/PLY_(file_format)
[8] Khronos Group — glTF 2.0 specification: https://registry.khronos.org/glTF/specs/2.0/glTF-2.0.html
[9] Apple — AR Quick Look: https://developer.apple.com/augmented-reality/quick-look/
[10] three.js — documentation: https://threejs.org/docs/
[11] Prusa Knowledge Base — Modeling with 3D printing in mind: https://help.prusa3d.com/article/modeling-with-3d-printing-in-mind_164135
[12] Prusa Blog — lithophanes and filament transparency: https://blog.prusa3d.com/lets-play-with-filament-transparency-how-to-make-amazing-lifelike-color-photos-with-just-a-simple-color-change_104338/
[13] Polymers (2024) 16(20):2862 — FFF parameters and translucency of PLA: https://doi.org/10.3390/polym16202862
[14] Wikipedia — Lithophane: https://en.wikipedia.org/wiki/Lithophane
[15] W3C — CSS Values and Units Level 3 (absolute lengths): https://www.w3.org/TR/css-values-3/
[16] Adobe Substance 3D — OpenGL vs DirectX normal maps: https://experienceleague.adobe.com/en/docs/substance-3d/bakers/common-questions/what-is-the-difference-between-the-opengl-and-directx-normal-format
[17] Blender Manual — Normal Map node: https://docs.blender.org/manual/en/latest/render/shader_nodes/displacement/normal_map.html
[18] Wikipedia — Normal mapping: https://en.wikipedia.org/wiki/Normal_mapping
[19] SIL Open Font License: https://openfontlicense.org/
[20] Meshy Blog — Meshy 7.1 launch (10 Sep 2026): https://www.meshy.ai/blog/meshy-7-1-launch
[21] Meshy Blog — Meshy 7 and image-to-3D geometry alignment: https://www.meshy.ai/blog/meshy-7-image-to-3d-geometry-alignment
[22] Meshy Help — Meshy 7 vs Meshy 6 vs Meshy 6 Lite: https://help.meshy.ai/en/articles/15697060-meshy-7-vs-meshy-6-vs-meshy-6-lite-which-ai-model-should-you-use
[23] Tripo Developer Docs — Tripo H3.1 (v3.1-20260211): https://developers.tripo3d.ai/en/models/v3-1
[24] VoxelMatters — Tripo AI launches P2.0 with native quad meshes (Sep 2026): https://www.voxelmatters.com/tripo-ai-launches-p2-0-model-to-generate-native-quad-meshes-for-production-pipelines/
[25] Hyper3D — Rodin Gen-2.5 launch press release (May 2026): https://pressadvantage.com/story/94732-hyper3d-launches-rodin-gen-2-5-bringing-sculpt-level-detail-and-production-control-to-ai-3d-generati
[26] arXiv — BANG: Dividing 3D Assets via Generative Exploded Dynamics (SIGGRAPH 2025): https://arxiv.org/abs/2507.21493
[27] befores & afters — Deemos launches Rodin Gen-2 (Oct 2025): https://beforesandafters.com/2025/10/02/deemos-launches-rodin-gen-2-a-groundbreaking-generative-ai-for-intuitive-3d-creation/
[28] GitHub — Tencent Hunyuan3D 2.1 (open weights): https://github.com/Tencent-Hunyuan/Hunyuan3D-2.1
[29] Hugging Face — Tencent Hunyuan 3D 2.1 Community License: https://huggingface.co/tencent/Hunyuan3D-2.1/blob/main/LICENSE
[30] Tencent Cloud — Hunyuan 3D product changelog: https://cloud.tencent.com/document/product/1804/120694
[31] CNBC — OpenAI rolls out GPT-6 Astra (3 Sep 2026): https://www.cnbc.com/2026/09/03/open-ai-astra-gpt-6-cyber.html
[32] AWS Bedrock — OpenAI GPT-6 Astra model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html
[33] VoxelMatters — GPT-6 Astra can turn photos into 3D models: https://www.voxelmatters.com/gpt-6-astra-can-turn-photos-into-3d-models-and-playable-environments/
[34] WaveSpeedAI — GPT-6 Astra image-to-3D endpoint: https://wavespeed.ai/models/openai/gpt-6-astra/image-to-3d
[35] WaveSpeedAI — Hunyuan 3D 3.1 image-to-3D Rapid endpoint: https://wavespeed.ai/models/wavespeed-ai/hunyuan-3d-v3.1/image-to-3d-rapid

## FAQ
Q: Is GPT-6 Astra a 3D model?
A: Not a dedicated one. GPT-6 Astra is OpenAI's general model; it creates 3D by planning and writing geometry. Through WaveSpeed it's offered as a 3D job that returns a GLB, which is how the studio runs it.

Q: When was GPT-6 Astra released?
A: OpenAI began rolling it out on 3 September 2026.

Q: How long does an Astra 3D job take?
A: Longer than other engines. Our test runs took between about 3.5 and 9 minutes; complex jobs can take more.

Q: Can GPT-6 Astra make models for Blender?
A: Yes — download the GLB and import it into Blender (File → Import → glTF 2.0).

Q: Does it make CAD files?
A: Press reports describe Astra writing CAD code in OpenAI's demos; in the studio the output is a GLB mesh, not a CAD file.

Q: Can I use it with ChatGPT?
A: The studio runs Astra for you — no ChatGPT or OpenAI account needed. Generation is part of every paid plan and credit pack.

Q: Is likeness to my photos exact?
A: Not guaranteed. Several clear photos and a precise brief give the closest result.

## Related pages
- Meshy AI: https://kenerateai.com/model/meshy-ai
- Tripo AI: https://kenerateai.com/model/tripo-ai
- Rodin AI: https://kenerateai.com/model/rodin-ai
- Hunyuan 3D: https://kenerateai.com/model/hunyuan-3d

## Sources
- Wikipedia — STL (file format): https://en.wikipedia.org/wiki/STL_(file_format)
- Wikipedia — Wavefront .obj file: https://en.wikipedia.org/wiki/Wavefront_.obj_file
- Wikipedia — FBX: https://en.wikipedia.org/wiki/FBX
- Khronos Group — glTF: https://www.khronos.org/gltf/
- 3MF Consortium — 3MF specification: https://3mf.io/specification/
- Wikipedia — Universal Scene Description (USD / USDZ): https://en.wikipedia.org/wiki/Universal_Scene_Description
- Wikipedia — PLY (file format): https://en.wikipedia.org/wiki/PLY_(file_format)
- Khronos Group — glTF 2.0 specification: https://registry.khronos.org/glTF/specs/2.0/glTF-2.0.html
- Apple — AR Quick Look: https://developer.apple.com/augmented-reality/quick-look/
- three.js — documentation: https://threejs.org/docs/
- Prusa Knowledge Base — Modeling with 3D printing in mind: https://help.prusa3d.com/article/modeling-with-3d-printing-in-mind_164135
- Prusa Blog — lithophanes and filament transparency: https://blog.prusa3d.com/lets-play-with-filament-transparency-how-to-make-amazing-lifelike-color-photos-with-just-a-simple-color-change_104338/
- Polymers (2024) 16(20):2862 — FFF parameters and translucency of PLA: https://doi.org/10.3390/polym16202862
- Wikipedia — Lithophane: https://en.wikipedia.org/wiki/Lithophane
- W3C — CSS Values and Units Level 3 (absolute lengths): https://www.w3.org/TR/css-values-3/
- Adobe Substance 3D — OpenGL vs DirectX normal maps: https://experienceleague.adobe.com/en/docs/substance-3d/bakers/common-questions/what-is-the-difference-between-the-opengl-and-directx-normal-format
- Blender Manual — Normal Map node: https://docs.blender.org/manual/en/latest/render/shader_nodes/displacement/normal_map.html
- Wikipedia — Normal mapping: https://en.wikipedia.org/wiki/Normal_mapping
- SIL Open Font License: https://openfontlicense.org/
- Meshy Blog — Meshy 7.1 launch (10 Sep 2026): https://www.meshy.ai/blog/meshy-7-1-launch
- Meshy Blog — Meshy 7 and image-to-3D geometry alignment: https://www.meshy.ai/blog/meshy-7-image-to-3d-geometry-alignment
- Meshy Help — Meshy 7 vs Meshy 6 vs Meshy 6 Lite: https://help.meshy.ai/en/articles/15697060-meshy-7-vs-meshy-6-vs-meshy-6-lite-which-ai-model-should-you-use
- Tripo Developer Docs — Tripo H3.1 (v3.1-20260211): https://developers.tripo3d.ai/en/models/v3-1
- VoxelMatters — Tripo AI launches P2.0 with native quad meshes (Sep 2026): https://www.voxelmatters.com/tripo-ai-launches-p2-0-model-to-generate-native-quad-meshes-for-production-pipelines/
- Hyper3D — Rodin Gen-2.5 launch press release (May 2026): https://pressadvantage.com/story/94732-hyper3d-launches-rodin-gen-2-5-bringing-sculpt-level-detail-and-production-control-to-ai-3d-generati
- arXiv — BANG: Dividing 3D Assets via Generative Exploded Dynamics (SIGGRAPH 2025): https://arxiv.org/abs/2507.21493
- befores & afters — Deemos launches Rodin Gen-2 (Oct 2025): https://beforesandafters.com/2025/10/02/deemos-launches-rodin-gen-2-a-groundbreaking-generative-ai-for-intuitive-3d-creation/
- GitHub — Tencent Hunyuan3D 2.1 (open weights): https://github.com/Tencent-Hunyuan/Hunyuan3D-2.1
- Hugging Face — Tencent Hunyuan 3D 2.1 Community License: https://huggingface.co/tencent/Hunyuan3D-2.1/blob/main/LICENSE
- Tencent Cloud — Hunyuan 3D product changelog: https://cloud.tencent.com/document/product/1804/120694
- CNBC — OpenAI rolls out GPT-6 Astra (3 Sep 2026): https://www.cnbc.com/2026/09/03/open-ai-astra-gpt-6-cyber.html
- AWS Bedrock — OpenAI GPT-6 Astra model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html
- VoxelMatters — GPT-6 Astra can turn photos into 3D models: https://www.voxelmatters.com/gpt-6-astra-can-turn-photos-into-3d-models-and-playable-environments/
- WaveSpeedAI — GPT-6 Astra image-to-3D endpoint: https://wavespeed.ai/models/openai/gpt-6-astra/image-to-3d
- WaveSpeedAI — Hunyuan 3D 3.1 image-to-3D Rapid endpoint: https://wavespeed.ai/models/wavespeed-ai/hunyuan-3d-v3.1/image-to-3d-rapid