Leaderboard / Model
GPT-6 Luna
by OpenAIAssistantChatGPT Work and Codex on paid plans, desktop app on Free and Go; API $0.10 / $0.50 per million input / output tokensUpdated 27 Sept 2026
GPT-6 Luna is OpenAI's most efficient GPT-6 model, released September 22, 2026 at $0.10 / $0.50 per million tokens. In Blender it writes Blender Python or drives your scene over MCP. It entered Season 3 at xhigh reasoning effort, so the blind votes show how a model at a twentieth of Sol's price holds up in 3D against Sol, Astra and the rest.
How to use GPT-6 Luna with Blender
Ask it for a bpy script and run it in Blender's Scripting workspace, or connect Codex (or another MCP-capable client) running gpt-6-luna to MCP for Blender for live scene access.
Best for: High-volume or budget scripting work: many small bpy jobs at a very low per-token price. The blind votes will show how its 3D output holds up against the bigger models.
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-6 Luna 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 3Calibrating
- 1
- Entrants
- 12
- Builds
- of 12 prompts
- 28
- Times voted on
- GPT-6 LunaCalibrating1389 ±828–17–1
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 GPT-6 Luna?
GPT-6 Luna is OpenAI's most efficient GPT-6 model, released on September 22, 2026 alongside GPT-6 Sol and trained with methods similar to GPT-6 Astra. OpenAI describes it as its most efficient model for focused, high-volume tasks. It has a 1.05M-token context window, up to 128K output tokens, a May 18, 2026 knowledge cutoff, and reasoning effort from none up to max (default medium).
On the API (model ID gpt-6-luna) it costs $0.10 per million input tokens, $0.01 cached and $0.50 output, down from $0.20 / $1.20 for GPT-5.6 Luna. It is in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise and Edu, and in the desktop app for Free and Go users.
It has no first-party Blender integration: it writes Blender Python (bpy) you run in the Scripting workspace, or drives a live scene through an MCP-capable client such as Codex connected to MCP for Blender.
Luna joined Season 3 on September 27, 2026, at xhigh reasoning effort, writing a single bpy script per prompt executed once in headless Blender 5.1. All twelve prompts produced geometry, with no script errors.
GPT-6 Luna at a glance
- Released: September 22, 2026
- API model ID: gpt-6-luna
- Price: $0.10 per million input tokens, $0.01 cached input, $0.50 output (GPT-5.6 Luna: $0.20 / $1.20)
- Context window: 1.05M tokens; max output: 128K tokens
- Reasoning effort: none, low, medium (default), high, xhigh, max
- Knowledge cutoff: May 18, 2026
- Available in ChatGPT Work and Codex on paid plans, the desktop app for Free and Go users, and the API
GPT-6 Luna vs GPT-6 Sol
Luna costs a twentieth of Sol per token. OpenAI reports it at 66.6% on DeepSWE v1.1 at max effort against Sol's 68.8%, and says Luna at max beats GPT-5.6 Sol at medium on OSWorld 2.0 offline at a tenth of the cost.
None of those measure 3D. Luna ran the same twelve prompts as Sol under the same rule, one script per prompt executed once in headless Blender 5.1, at the same xhigh effort, so the blind votes show how much of Sol's 3D quality survives at Luna's price.
Arena status
GPT-6 Luna joined Season 3 on September 27, 2026, at xhigh reasoning effort, the same setting as GPT-6 Sol, Claude Opus 5.5 and Grok 4.7. All twelve prompts produced geometry with no script errors, and every output is on this page exactly as generated. It entered blind voting at the Glicko-2 default rating and stays unranked until it clears the calibration floors.
Strengths and limitations
Strengths
- Strength:By far the cheapest GPT-6 model: $0.10 / $0.50 per million tokens, a twentieth of GPT-6 Sol's price
- Strength:OpenAI reports it at 66.6% on DeepSWE v1.1 at max effort, close to Sol's 68.8%
- Strength:1.05M-token context and 128K output tokens, room for long bpy scripts
- Strength:Produced a complete twelve-prompt arena set at xhigh effort, every output with real geometry
Limitations
- Limitation:No blind-vote record yet: it joined Season 3 on September 27, 2026, so its rating is uncalibrated and nothing here is a quality claim
- Limitation:OpenAI positions it for focused, high-volume tasks; GPT-6 Astra remains its strongest model
- Limitation:Not a mesh generator: geometry comes from code, which makes organic and sculpted subjects the structural weak spot
- Limitation:No official Blender connector; integration is via MCP tooling or pasted scripts
GPT-6 Luna and Blender: FAQ
When was GPT-6 Luna released?
OpenAI released GPT-6 Luna on September 22, 2026, together with GPT-6 Sol. It is in ChatGPT Work and Codex on paid plans, in the desktop app for Free and Go users, and in the API as gpt-6-luna.
How much does GPT-6 Luna cost?
$0.10 per million input tokens, $0.01 for cached input and $0.50 per million output tokens, down from $0.20 / $1.20 for GPT-5.6 Luna.
Can GPT-6 Luna use Blender?
Yes: it writes Blender Python you run in the Scripting workspace, or Codex (or another MCP-capable client) running it connects to MCP for Blender and works in a live scene. There is no official OpenAI Blender connector.
Is GPT-6 Luna good enough for 3D modelling?
Not settled yet. Its xhigh run is in blind voting against GPT-6 Sol, GPT-6 Astra, Claude Opus 5.5 and the rest on the same twelve prompts, all twelve with geometry.