Measured harness ledgerPublic result
GLM 5.3Off-road Mud Game — GLM 5.3 Max · Attempt 2
Build a one-request procedural off-road driving game with vehicle physics, independent suspension, streamed terrain, mud feedback, and multiple cameras.
Max reasoningHeadline result
- Workflow cost
- $9.19
- Wall-clock
- 7369.0s wall-clock
- Processed tokens
- 38.26M processed
- Record state
- partial_runtime_functional_visual_fail_metrics_unavailable
Public summary
GLM 5.3 Max partial_runtime_functional_visual_fail_metrics_unavailable ledger: 7369.0s wall-clock, 38.26M processed, and $9.19 OpenCode/OpenRouter recorded usage estimate, not a subscription cash-charge invoice.
Run identity and stack
- Result ID: off-road-driving-game-threejs-glm-5.3-max-openrouter-attempt-2
- Technical model: GLM 5.3
- Provider: OpenRouter
- Client: OpenCode
- Attempt: 2
- Stack: OpenRouter
- Stack: OpenCode
- Stack: OpenCode → OpenRouter
- Stack: Technical model/configuration: GLM 5.3
- Stack: Attempt 2
- Stack: Three.js vehicle workflow
- Stack: Harness v1 off-road prompt
- Stack: Requested tool profile: OpenCode CLI
Primary artifact integrity
- Kind: manifest-verified-artifact
- Path: artifacts/off-road-driving-game-threejs-glm-5.3-max-openrouter-attempt-2/source/src/main.js
- SHA-256: 7ed01d67556a4b3cde934f3ccbaf42a747d51241c10d971705e60350bb415c5e
Validation evidence
- Result: PASS
- Path: artifacts/off-road-driving-game-threejs-glm-5.3-max-openrouter-attempt-2/validation/audio-test.public.json
- SHA-256: 9d7009860f95a420809e15c2ee4b2b9aefa2eaec1467570dc737ca1a9b29ffff
Recorded caveats
- The final planned visual-judge call returned HTTP 402 insufficient OpenRouter credits; no missing judge output is fabricated.
- The archived 300-second receipt was generated before the final 1024m floating-origin threshold edit, so origin rebasing is not claimed as endurance-tested.
- The visual-quality gate fails: the captures show sparse/placeholder-like forest dressing, visible terrain repetition or boundaries, and clipped cockpit framing.
- The OpenCode/OpenRouter cost is provider usage accounting, not a subscription cash invoice.
- No RTX Pro 6000 render or blind-evaluation record is archived.
Visible evidence gaps
- post-generation operator metrics receipt
- final visual-judge round
- floating-origin rebasing in the endurance receipt
- photoreal visual-quality pass
- RTX Pro 6000 final render
- 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.