Measured harness ledgerPublic result
GPT-6 AstraMacBook-class cinematic ad scene — GPT-6 Astra Max
Create one polished MacBook-class product-ad shot in Blender with a modeled device, legible industrial detail, intentional materials, lighting, camera movement, and a validator-ready scene.
Max reasoningHeadline result
- Workflow cost
- $20.35
- Wall-clock
- 65m 04s (qualified snapshot span) wall-clock
- Processed tokens
- 11.79M processed
- Record state
- partial_post_task_metrics_snapshot_with_independent_validation
Public summary
GPT-6 Astra Max partial_post_task_metrics_snapshot_with_independent_validation ledger: 65m 04s (qualified snapshot span) wall-clock, 11.79M processed, and $20.35 API-equivalent estimate from a qualified post-task snapshot, not a subscription invoice.
Run identity and stack
- Result ID: macbook-cinematic-blender-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: Cinematic product scene
- Stack: Harness v1 MacBook validator
- Stack: Requested tool profile: blender-mcp
Primary artifact integrity
- Kind: blender-cinematic-product-ad-scene
- Path: artifacts/macbook-cinematic-blender-gpt-6-astra-max/macbook_cinematic_final.blend
- SHA-256: 9f8bba5b2c1e9d4c5b08d57540da4fc98c886cd59df7ac4c90716e31103f78d6
Validation evidence
- Result: 49/49 checks recorded
- Path: artifacts/macbook-cinematic-blender-gpt-6-astra-max/validation_log.json
- SHA-256: e6ed20d8a1d5d15955a4c3f8c2d6ed409c09f1e28bc43a536a86d0ad4cd78ec6
- Validator SHA-256: 54591de8d4ef2953c6f7e2fe3e4f4b3d9407b953049cc858b1012afaf9424d51
Recorded caveats
- The supplied metrics snapshot includes the beginning of a later metrics follow-up, so its timing, token total, and cost are not a clean completed-run score.
- The raw parser did not recognize user-message payloads; the displayed 65m 04s span is a schema-aware derivation and user-turn count remains unavailable.
- Wall-clock includes tools and idle gaps, not just model compute.
- Output tokens include hidden reasoning, visible prose/code, and tool-call JSON.
- Cache reads are discounted, so processed-token volume overstates cost.
- Cache-creation tokens are unavailable in the transcript schema; zero is not proof of no billable cache writes.
- The canonical validator and headless reopen confirm configured scene properties, not visual quality, final-animation rendering, render performance, or retained Blender MCP tool use.
- Blind evaluation remains outstanding.
Visible evidence gaps
- clean completed-run token/timing receipt
- benchmark user-turn count
- Blender MCP tool-use transcript or disclosure
- render-hardware 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.