Measured harness ledgerPublic result
DeepSeek V4 FlashInteractive Three.js voxel map viewer — DeepSeek V4 Flash Max
Build a Three.js voxel map viewer with selectable Jungle Temple, Oasis Outpost, and Shrine Village maps, camera reset behavior, and OrbitControls verification.
Max reasoningHeadline result
- Workflow cost
- $0.03
- Wall-clock
- 37m 33.5s wall-clock
- Processed tokens
- 8.48M processed
- Record state
- partial_token_timing_source_and_static_smoke_ledger
Public summary
DeepSeek V4 Flash Max partial_token_timing_source_and_static_smoke_ledger ledger: 37m 33.5s wall-clock, 8.48M processed, and $0.03 OpenCode Go API-equivalent usage accounting, not an itemized subscription cash charge.
Run identity and stack
- Result ID: interactive-voxel-map-viewer-threejs-deepseek-v4-flash-max
- Technical model: DeepSeek V4 Flash
- Provider: OpenCode Go
- Stack: OpenCode Go
- Stack: Technical model/configuration: DeepSeek V4 Flash
- Stack: Three.js voxel viewer
- Stack: OrbitControls
- Stack: Harness v1 voxel-viewer prompt
Primary artifact integrity
- Kind: interactive-threejs-voxel-map-viewer
- Path: artifacts/interactive-voxel-map-viewer-threejs-deepseek-v4-flash-max/source/index.html
- SHA-256: e27ac9d8b91aeda0020438f2d7f808d37818125ab04714e564db81d3672527c3
Recorded caveats
- Wall-clock is end-to-end workflow latency including tool execution, installs, browser checks, and idle gaps; not model-only compute.
- One OpenCode agent turn can contain many billable model requests.
- Visible output and reasoning are separately metered and both output-priced in this accounting.
- Cache reads are deeply discounted, so total processed tokens overstate cost.
- The public source was smoke-tested but not independently browser-run during archiving. No local FPS, hardware, final capture, or blind-evaluation record is supplied.
- The source receipt records two benchmark user turns, so this is a multi-user-turn protocol deviation rather than a one-shot run.
Visible evidence gaps
- one-user-turn benchmark protocol
- browser runtime verification of all map switches and OrbitControls
- browser and hardware environment
- local FPS measurement
- final capture
- 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.