~ same prompt, different LLM — every model gets the same brief in Blender, you judge the render ~
OpenAI / GPT 6 Astra / thinking:max has entered the arena NEW!
Render Window — hero
hero render
camera + framing set up by the AI contestant
3D Viewport — abandoned-village User Persp
LOADING 3D VIEW
$0.02Cost
91sDuration
12Tool calls
13Turns
Blender 5.1.0MCP 1.22026-07-15
Tokens: 83402/2004Vision: Votes W-L-T: 1-5-0DONE ✓
B-BENCH — TRANSCRIPT Google / Gemini 3.1 Flash Lite · r1 · 2026-07-15
RUN LOG — abandoned-village · Google / Gemini 3.1 Flash Lite · r1 · ok
SYSTEM PROMPTthe control — every model on this task got these instructions
PROMPTblender-bench v7HARNESSraw-v2
SHOW FULL PROMPT · 2016 chars
You are participating in Blender-Bench, a public benchmark where AI models complete the same Blender task.

Your result will be compared head-to-head with another model. Anyone on the internet can vote, based only on the published visual outputs they can inspect: the interactive 3D scene and rendered presentation.

Complete the request as well as you can. The benchmark does not prescribe a workflow or visual style; use whichever Blender techniques you consider appropriate. Make the important requirements evident in the published result, save a complete scene, and present the work clearly in every requested output.

The final published capture is rendered by the benchmark from your saved scene using Cycles at 1536 x 1536, 64 samples, denoising enabled, and GPU acceleration when available with a CPU fallback. The saved scene is also exported to GLB with geometry, materials, the contestant camera, and supported lights for interactive inspection. Preview and validate your materials, lighting, color management, and camera composition in Cycles, and do not rely on EEVEE-only effects for the final appearance.

## Harness Instructions

You are operating Blender through the provided tools. The scene starts empty. Build what the user's prompt asks for in the current scene. Work step by step and verify your own work. Before you finish, set up a camera and lighting to frame a single, well-composed presentation image of your result — position and aim the camera, and add or adjust lights so the subject reads clearly. The final benchmark capture overrides the scene render settings and renders in Cycles at 1536x1536, 64 samples, with denoising and GPU acceleration (CPU fallback). Preview and validate materials, lighting, color management, and composition in Cycles; do not rely on EEVEE-only effects. The saved scene is also exported to GLB with geometry, materials, your camera, and supported lights for interactive inspection. When you are finished, reply with the single word DONE and no tool calls.
--TASK abandoned-village: ## Task Complete the following Blender task: <task> An abandoned village slowly reclaimed by nature. Weathered houses with rusted, corroded metal roofs; overgrown foliage creeping up the walls and through broken windows; moss, grass, and vines spreading across the ground and over the rooftops. Convey age, decay, and the quiet return of the wild. </task>
27:24>>execute_blender_code({"user_prompt":"Initial scene setup for an abandoned village house.","code":"import bpy\n\n# Clear existing objects\nbpy.ops.object.select_all(action='SELECT')\nbpy.ops.object.delete()\n\n# Create a ground plane\nbpy.ops.mesh.primitive_plane_add(size=20, location=(0, 0, 0))\nground = bpy.context.active_object\nground.name = \"Ground\"\n\n# Create a simple house base\nbpy.ops.mesh.primitive_cube_ad…)
27:24<<Code executed successfully:
27:27>>search_sketchfab_models({"query":"abandoned house ruins overgrown foliage"})
27:27<<Error searching Sketchfab models: Communication error with Blender: Unknown command type: search_sketchfab_models
27:29>>get_sketchfab_status({})
27:29<<Sketchfab integration is currently disabled. To enable it: 1. In the 3D Viewport, find the BlenderMCP panel in the sidebar (press N if hidden) 2. Check the 'Use assets from Sketchfab' checkbox 3. Enter your Sketchfab API Key 4. Restart the connection to Claude
27:31>>generate_hyper3d_model_via_text({"text_prompt":"weathered abandoned small wooden house with a rusted metal roof, overgrown with moss and vines, highly detailed, realistic style"})
27:31<<Error generating Hyper3D task: Communication error with Blender: Unknown command type: create_rodin_job
27:34>>get_hyper3d_status({})
27:34<<Hyper3D Rodin integration is currently disabled. To enable it: 1. In the 3D Viewport, find the BlenderMCP panel in the sidebar (press N if hidden) 2. Check the 'Use Hyper3D Rodin 3D model generation' checkbox 3. Restart the connection to Claude
27:36>>get_polyhaven_status({})
27:36<<PolyHaven integration is currently disabled. To enable it: 1. In the 3D Viewport, find the BlenderMCP panel in the sidebar (press N if hidden) 2. Check the 'Use assets from Poly Haven' checkbox 3. Restart the connection to Claude
27:41>>execute_blender_code({"code":"import bpy\nimport random\n\n# Add some procedural details to the scene\ndef create_mossy_patch(location):\n bpy.ops.mesh.primitive_ico_sphere_add(subdivisions=2, radius=0.5, location=location)\n obj = bpy.context.active_object\n obj.scale = (random.uniform(1, 3), random.uniform(1, 3), 0.5)\n return obj\n\n# Scatter some \"foliage/moss\" patches\nfor i in range(10):\n loc =…)
27:41<<Code executed successfully:
27:44>>get_viewport_screenshot({"user_prompt":"Checking the current scene state."})
27:45<<
27:45imgimage delivered — get_viewport_screenshot
27:52ai>thought
27:52>>execute_blender_code({"code":"import bpy\n\n# Add some randomness to the terrain to look more \"reclaimed\"\ndef deform_ground():\n ground = bpy.data.objects[\"Ground\"]\n if \"subsurf\" not in [m.name for m in ground.modifiers]:\n mod = ground.modifiers.new(name=\"subsurf\", type='SUBSURF')\n mod.levels = 3\n mod.render_levels = 3\n \n # We could add a displace modifier here with a cl…)
27:52<<Code executed successfully:
27:58>>get_viewport_screenshot({"user_prompt":"Previewing the final setup."})
27:59<<
27:59imgimage delivered — get_viewport_screenshot
28:06>>execute_blender_code({"code":"import bpy\n\n# Add a simple environment texture/light to make it look moody\nworld = bpy.context.scene.world\nif not world:\n world = bpy.data.worlds.new(\"World\")\n bpy.context.scene.world = world\n\nworld.use_nodes = True\nnodes = world.node_tree.nodes\nnodes.clear()\n\n# Create a dark, moody background\nbg = nodes.new('ShaderNodeBackground')\nbg.inputs['Color'].default_value = …)
28:06<<Error executing code: Communication error with Blender: Code execution error: 'bpy_prop_collection[key]: key "Color" not found'
28:12>>execute_blender_code({"code":"import bpy\n\n# Re-try adjusting world lighting carefully\nworld = bpy.context.scene.world\nif not world:\n world = bpy.data.worlds.new(\"World\")\n bpy.context.scene.world = world\n\nworld.use_nodes = True\nnodes = world.node_tree.nodes\n# Keep the existing nodes if any, just modify\nif \"Background\" in nodes:\n bg = nodes[\"Background\"]\nelse:\n bg = nodes.new('ShaderNodeB…)
28:12<<Code executed successfully:
28:17ai>DONE
28:17--status: ok · DONE declared
>
FULL CALL INSPECTOR
OPEN RAW JSONL ↗
0 SHOWN
LOADING COMPLETE TRANSCRIPT…
Votes cast on this bench: 000482 · 157 runs · 25 contestants
Best viewed with Netscape Navigator 4.07 at 800×600 · real runs, live ratings