Measured harness ledgerPublic result
GPT-6 AstraHigh-voxel-density Shrine Village diorama — GPT-6 Astra Max
Create a complete inspectable Blender voxel diorama of a shrine village at a fixed 0.05-unit voxel resolution through the Blender MCP workflow.
Max reasoningHeadline result
- Workflow cost
- $9.49
- Wall-clock
- 35m 39.2s wall-clock
- Processed tokens
- 5.89M processed
- Record state
- completed_generation_turn_with_schema_caveats_and_unverified_render_acceptance
Public summary
GPT-6 Astra Max completed_generation_turn_with_schema_caveats_and_unverified_render_acceptance ledger: 35m 39.2s wall-clock, 5.89M processed, and $9.49 API-equivalent estimate from the completed-generation record, not a subscription invoice.
Run identity and stack
- Result ID: shrine-village-gpt-6-astra-max
- Technical model: gpt-6-astra
- Provider: OpenAI Codex
- Stack: OpenAI Codex
- Stack: Technical model/configuration: gpt-6-astra
- Stack: Blender MCP
- Stack: 0.05-unit voxel grid
- Stack: Inspectable .blend artifact
- Stack: Requested tool profile: blender-mcp
Primary artifact integrity
- Kind: blender-scene
- Path: artifacts/shrine-village-gpt-6-astra-max/Shrine_Village.blend
- SHA-256: e33bd64cb1e5d8d72084d1d9513f486eb87b06ef03a892674d9ea8ceb0df5855
Recorded caveats
- The raw parser did not recognize the source user-message representation, so timing is reconstructed from the frozen transcript's task-complete marker rather than a parser user timestamp.
- Wall-clock is end-to-end elapsed time, including tools and idle gaps, not model-only compute.
- Output tokens include hidden reasoning, visible prose/code, and tool-call JSON.
- Cache reads are discounted, so total processed tokens overstate cost.
- The grid audit and viewport/clay review are model-supplied evidence. The archive operator independently reopened the scene and checked structure, but did not rerender it or perform a fresh visual judge loop.
- Final Cycles-render acceptance, render-performance measurement, and blind evaluation remain unverified.
Visible evidence gaps
- recorded Blender MCP tool-use transcript or explicit tool-use disclosure
- final render capture with renderer and hardware record
- render or real-time performance measurement
- blind-evaluation record
Public result only
This result keeps its public summary and evidence, but it does not currently have a matching Builder test with prompts, projects, Harness workflows or skills.