Fighter Game v1 — Claude Opus 5 Max

Measured harness ledgerPublic result

Fighter Game v1 — Claude Opus 5 Max

Improve the complete arena environment of a fixed Three.js fighting game through Blender MCP without changing the fighters, combat, HUD, effects, or audio.

Max reasoningHeadline result

Workflow cost: ≥$104.53
Wall-clock: 1h 23m 18.7s wall-clock
Processed tokens: 138.89M processed
Record state: artifact_independent_boot_metrics_ledger_no_blind_evaluation

Public summary

Claude Opus 5 Max artifact_independent_boot_metrics_ledger_no_blind_evaluation ledger: 1h 23m 18.7s wall-clock, 138.89M processed, and ≥$104.53 First-party Claude API list-price equivalent for the measured token mix; not an itemized subscription charge (lower bound).

Run identity and stack

Result ID: courtyard-arena-blender-mcp-opus-5-max
Technical model: claude-opus-5
Provider: Provider not separately recorded
Client: Claude Code
Stack: Claude Code
Stack: Technical model/configuration: claude-opus-5
Stack: Blender MCP
Stack: Three.js fighting-game fixture
Stack: Harness v1 frozen prompt

Cost basis

The token snapshot was taken while the session was still growing (401 of an eventual 409 assistant turns).
Transcript usage.cache_creation reports ephemeral_1h_input_tokens = 2,182,711 and ephemeral_5m_input_tokens = 0, i.e. 100% one-hour writes, billed at 2x base input.

Primary artifact integrity

Kind: manifest-verified-artifact
Path: artifacts/courtyard-arena-blender-mcp-opus-5-max/source/web/models/arena.glb
SHA-256: 4833f0391ff09b1240aaffcd4d7b4e101e051385ed7bfe0b79bb9b9d582d5e23

Validation evidence

Result: PASS

Recorded caveats

Wall-clock is end-to-end workflow latency including local Blender builds, texture generation, and browser captures — not model-only compute.
Output tokens include hidden reasoning, code, and tool-call JSON.
Cache reads bill at 0.1x input, so the 138.9M total processed figure massively overstates cost.
Every token, timing, and cost figure is a lower bound: the transcript grew from 401 to 409 assistant turns while metrics were being collected.
This is an environment-improvement task over a fixed starter, so its token and cost figures are not comparable to the archive's from-scratch build rows.
The archive is a delta over the committed fixture, not a standalone runnable tree; overlay it on benchmarks/fixtures/courtyard-arena-blender-mcp.v1/starter/ to reproduce the delivered copy.
No frame-rate receipt exists — neither the model nor the operator recorded a valid FPS sample.
No blind evaluation and no side-by-side against the reference run are recorded; the fixture's judging anchors were not scored.

Visible evidence gaps

blind-evaluation record
standardized FPS and hitch receipt under sustained play
scored side-by-side against the reference run's judging anchors

Builder test available

This result is part of a Builder test. Open it for the exact prompt and any released projects, RemakeBench Harness workflows and production skills. Public proof and known evidence gaps stay visible here.

Fighter Game v1 · Builder projects v1

View Fighter Game v1 · Builder projects v1 Return to Claude Opus 5