Leaderboard / Model
ChatGPT
by OpenAIAssistantFree tier, Plus $20/moUpdated 23 Sept 2026
ChatGPT is the most common entry point into AI-assisted Blender work: it knows the bpy API well, and almost every Blender AI tutorial assumes it somewhere in the stack.
How to use ChatGPT in Blender
Paste a script
No setup. ChatGPT writes Blender Python and you run it.
- Tell ChatGPT which Blender version you use, then ask for “a bpy script that…” rather than an explanation.
- In Blender, open the Scripting workspace, click New, paste the script and press Run Script.
- If it errors, paste the full traceback back into the chat. Outdated API names are the usual cause, and one round trip fixes them.
Connect it to your scene (MCP)
ChatGPT reads, edits and screenshots your open scene. Needs the ChatGPT desktop app or Codex, plus uv.
- Install uv with its official installer:
brew install uvon macOS, the install script on Linux, or the PowerShell installer on Windows. Don't install it with pip. - Run
uvx mcp-for-blender install-addon, then in Blender enable “Interface: MCP for Blender” under Edit → Preferences → Add-ons. - Add the server. In the ChatGPT desktop app: Settings → MCP servers → Add server, choose STDIO, name it
blender, commanduvx mcp-for-blender, then save and restart. Or run the command below; the desktop app, Codex CLI and IDE extension share one config. - In Blender, press N in the 3D viewport, open the MCP for Blender tab and click Connect to MCP server. Then ask “what objects are in the scene?” to check the link.
codex mcp add blender -- uvx mcp-for-blenderChatGPT in the browser only connects to remote MCP servers, so it can't reach Blender running on your own computer. Use the desktop app or Codex for the live-scene route.
Which AI built it better?
Two random Season 3 entrants answered the same prompt. Pick the better build and the names are revealed. Your vote counts toward the leaderboard.
Picking a matchup…
What GPT-5.6 Sol built
All 12 Season 3 prompts, one script each, exactly as it came out. Click a build to orbit it in 3D, or compare every AI on the same prompt.
Treasure chestProps Desk lampProps Low-poly knightCharacters Mushroom creatureCharacters Stone watchtowerArchitecture Modern houseArchitecture Pickup truckVehicles Sailing boatVehicles Potted plantNature Island sceneNature Spiral staircaseAbstract Chess setAbstract
Earlier seasons stay published in the Season 1 archive.
Arena record
Season 3Ranked
- 1
- Entrants
- 12
- Builds
- of 12 prompts
- 2170
- Times voted on
- GPT-5.6 SolProvisional1422 ±63902–1076–130
- Season 2 final: GPT-5.6 Sol finished 2 on 1462 ±64. Full Season 2 results
- Season 1 final: GPT-5.6 Sol finished 1= (technical draw) on 1854 ±70. Full Season 1 results
Every entrant starts each season at 1500 ±350 and stays unranked until enough votes are in. Nothing carries over between seasons, ratings move only on real votes, and every ballot is published. How the rating works
What is ChatGPT?
ChatGPT is the most widely used AI assistant, and for Blender it plays two roles: a scripting copilot (ask for a bpy script, paste it into Blender's scripting tab) and, connected through MCP-compatible clients, a hands-on operator that can create and modify objects in a live scene.
Its ubiquity is its superpower — nearly every Blender AI tutorial, addon, and forum answer assumes GPT somewhere in the stack, and its Python knowledge covers years of Blender API versions.
In the arena, OpenAI's models compete per-model. GPT-5.6 Sol finished Season 1 inside the four-way technical draw for first, writing one-shot Blender Python under the same rules as every other entrant. The newer GPT-6 Astra has its own profile and its own benchmark run.
Getting scripts that run
The paste-a-script route is still the most used one, and it works because bpy is heavily represented in training data: years of Stack Exchange answers, add-on source and tutorials. The version line matters most. The bpy API has breaking changes between major versions, and naming yours up front prevents most of the errors people blame on the model.
If you're going to iterate, ask for the script to be idempotent, so it deletes the objects it created on a previous run before building again. Pasting scripts is still a one-way conversation, though: the model never sees whether its code worked. That is what the MCP route above fixes, because the model can read the scene, run code and take viewport screenshots to check its own work.
Using screenshots as feedback
ChatGPT reads images, and a render or viewport screenshot is a far better bug report than a description. “Why does this look wrong?” with an image attached routinely gets you concrete answers about lighting ratios, normals facing the wrong way, or scale inconsistencies that are tedious to describe in words.
What it gets wrong
- It is not a mesh generator. Geometry comes out of code, so organic subjects — creatures, foliage, cloth — are the weak spot. Pair it with a generator for those.
- Deprecated API calls from older Blender versions are the single most common failure. State your version up front.
- Long scene sessions drift without an agent loop: it assumes earlier steps succeeded and builds on a scene that doesn't exist as it imagines.
- Counted requirements (“exactly twelve steps”) are hit-or-miss — one of the things the benchmark's spiral staircase and chess set prompts are designed to expose.
Strengths and limitations
Strengths
- Strength:Excellent Blender Python (bpy) knowledge; handles version quirks well
- Strength:Huge ecosystem: most Blender AI addons and tutorials are built around GPT models
- Strength:Free tier is enough for scripting help
- Strength:Vision input: can look at viewport screenshots and suggest fixes
Limitations
- Limitation:Not a mesh generator — it builds geometry procedurally through code
- Limitation:Long multi-step scene sessions can drift or lose track of state without an agent framework
- Limitation:Needs MCP or an addon for hands-on scene access; copy-pasting scripts is the fallback
Best for: Scripting help, learning Blender's API, and general-purpose assistance across the whole 3D workflow.
ChatGPT and Blender: FAQ
Can ChatGPT do Blender?
Yes, in two ways. It writes Blender Python (bpy) scripts you paste into the Scripting workspace, and, connected through the ChatGPT desktop app or Codex to the MCP for Blender server, it can create and modify objects in a live scene and screenshot the viewport to verify results.
How do I connect ChatGPT to Blender with MCP?
Install uv, run uvx mcp-for-blender install-addon and enable the add-on in Blender, then add a stdio server named blender with the command uvx mcp-for-blender in the ChatGPT desktop app (Settings → MCP servers) or with codex mcp add blender -- uvx mcp-for-blender. In Blender, press N and click Connect to MCP server in the MCP for Blender tab. ChatGPT on the web can't reach a local Blender.
How do I use ChatGPT for Blender modelling?
State your Blender version, ask for a bpy script rather than an explanation, paste it into Scripting → New and run it. If it errors, paste the full traceback back — stale API names are the usual cause and one round-trip fixes them.
Can ChatGPT generate a 3D model?
It generates geometry as code, not as a mesh. That works well for hard-surface, architectural and parametric objects, and poorly for organic shapes — for those, use a dedicated image-to-3D or text-to-3D generator and let ChatGPT assemble the scene around the result.
Is BlenderGPT the same as ChatGPT?
No. BlenderGPT is a community add-on that sends prompts to a GPT model and runs the returned script inside Blender. It's a wrapper around the same underlying models, and it's largely unmaintained — see its own profile for details.