Claude Haiku 5.5 (effort default (no --effort flag) · thinking on) on One-shot a Minecraft-style voxel game in three.js

AClaude Haiku 5.5default · thinking on · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-16 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first tryCompare

Estimated cost: $0.02
Wall time: 2m
Human step-ins: 0
Outcome: Pass

Compare with other models on this task →

Exact prompt

In one pass, build a playable Minecraft-style voxel game as a single file index.html using three.js loaded from https://unpkg.com/three@0.160.0/build/three.module.js (ES module import). Requirements: a generated terrain of at least 32x32 columns with at least 3 block types; first-person WASD movement with mouse look (pointer lock); left click breaks the targeted block, right click places a block; number keys 1-3 select block type; a crosshair and a small HUD showing the selected block. No other files.

Task prompt given to the agent (headless, offline, synthetic fixture inputs): In one pass, build a playable Minecraft-style voxel game as a single file index.html using three.js loaded from https://unpkg.com/three@0.160.0/build/three.module.js (ES module import). Requirements: a generated terrain of at least 32x32 columns with at least 3 block types; first-person WASD movement with mouse look (pointer lock); left click breaks the targeted block, right click places a block; number keys 1-3 select block type; a crosshair and a small HUD showing the selected block. No other files. Inspired by https://x.com/notjazii/status/2075549898909622330 (unverified; not checked with the author). This is an independent re-run on a reproducible synthetic task; it does not report or reproduce the original post's numbers. Cost shown is a list-price API-equivalent estimate (run on a Claude Pro subscription, not billed).

Success criteria: All automated rubric checks pass: imports three@0.160.0 module; pointer lock; WASD; raycast break/place (mouse buttons 0 and 2); block select keys 1-3; crosshair + HUD elements; terrain >=32x32 (size constant); no uncaught console errors in headless Chrome.

Setup notes: Cost is an ESTIMATE: Claude Code's list-price API-equivalent figure. The run used a Claude Pro subscription, so nothing was billed. input_tokens = all input incl. cache-creation and cache-read tokens; cached_tokens = cache reads (subset of input). Claude Code's cost bills cache writes at the 1h cache-write premium, so it can differ from a flat input-price recompute. Web disabled; tools Read/Write/Edit/Glob/Grep/Bash; single run; scored automatically (8/8 checks). Case 13 was rescored after the run (no rerun): three rubric checks (crosshair id, voxel raycast, 1-3 key handling) were widened after manual review found false negatives; final score is the rescored one.

Estimated cost List-price estimate; not actually billed.

Metrics

Estimated cost: $0.02
Wall time: 2m
Human step-ins: 0
Turns: 6

Model usage
Model	Input tok	Output tok	Cost
AClaude Haiku 5.5anthropic/claude-haiku-5.5effort default (no --effort flag) · thinking on	105,974	25,422	$0.02

Reproduce this run

Use the use case prompt on the use case page.
Same setup: Claude Haiku 5.5 (anthropic/claude-haiku-5.5, effort default (no --effort flag) · thinking on) via Claude Code 2.1.299 (headless -p), cloud.
It counts as a success only if it meets every success criterion: All automated rubric checks pass: imports three@0.160.0 module; pointer lock; WASD; raycast break/place (mouse buttons 0 and 2); block select keys 1-3; crosshair + HUD elements; terrain >=32x32 (size constant); no uncaught console errors in headless Chrome.
Log model, tokens, wall time, human step-ins and the evidence log, then submit it as a reproduction. A different model? Submit a head-to-head rerun.

How runs are scored: Methodology.

Was this useful?

⇄ Rerun on another model

Automated checks

Trust 80% comes from these checks (how Trust % works).

Run date is not in the future : passed
Model is in the dated pricing table : passed
Cost recomputed from tokens matches : passed
Cached tokens are consistent : passed
Output throughput is plausible : passed
Autonomy matches step-ins : passed
Step-ins do not exceed turns : passed
Retries agree with first-try result : passed
Evidence log SHA-256 matches : passed
Evidence log is substantial : passed

Evidence log

{
 "duration_api_ms": 102159,
 "stop_reason": "end_turn",
 "session_id": "[REDACTED]",
 "total_cost_usd": 0.01976502,
 "usage": {
  "input_tokens": 6,
  "cache_creation_input_tokens": 31546,
  "cache_read_input_tokens": 74422,
  "output_tokens": 25422,
  "output_tokens_details": {
   "thinking_tokens": 16211
  },
  "server_tool_use": {
   "web_search_requests": 0,
   "web_fetch_requests": 0
  },
  "service_tier": "standard",
  "cache_creation": {
   "ephemeral_1h_input_tokens": 31546,
   "ephemeral_5m_input_tokens": 0
  },
  "inference_geo": "global",
  "iterations": [
   {
    "input_tokens": 2,
    "output_tokens": 537,
    "cache_read_input_tokens": 43107,
    "cache_creation_input_tokens": 978,
    "cache_creation": {
     "ephemeral_5m_input_tokens": 0,
     "ephemeral_1h_input_tokens": 978
    },
    "type": "message"
   }
  ],
  "speed": "standard",
  "fallback_credit": null
 },
 "modelUsage": {
  "claude-haiku-5-5": {
   "inputTokens": 6,
   "outputTokens": 25422,
   "cacheReadInputTokens": 74422,
   "cacheCreationInputTokens": 31546,
   "webSearchRequests": 0,
   "costUSD": 0.01976502,
   "contextWindow": 1000000,
   "maxOutputTokens": 128000,
   "thinkingTokens": 16211,
   "canonicalModel": "claude-haiku-5-5",
   "provider": "firstParty",
   "costBasis": "list"
  }
 },
 "permission_denials": [],
 "terminal_reason": "completed",
 "fast_mode_state": "off",
 "fast_mode_disabled_reason": "sdk_opt_in_required",
 "subagent_stats": {
  "spawned": 0,
  "requested": {
   "background": 0,
   "foreground": 0,
   "unset": 0
  },
  "started_in_background": 0,
  "max_depth": 0,
  "spawned_by_subagents": 0,
  "completed": 0,
  "failed": 0,
  "killed": {
   "parent": 0,
   "user": 0,
   "system": 0
  },
  "refused": {
   "depth_limit": 0,
   "concurrency_limit": 0,
   "budget": 0
  },
  "by_type": {}
 },
 "safety_stops": 0,
 "is_error": false,
 "num_turns": 6,
 "subtype": "success",
 "api_error_status": null,
 "result": "I wrote the game as a single file at `./usecase-runner/runs/13-haiku/work/index.html`. The module script passes `node --check`, but I have not run it in a browser, so the rendering and pointer lock are untested.\n\nIt meets the requirements:\n\n- **Three.js:** imported from `https://unpkg.com/three@0.160.0/build/three.module.js`.\n- **Terrain:** a 40\u00d740 column world, 24 blocks tall, generated from a deterministic height function. It has three block types: grass (1), dirt (2) and stone (3).\n- **Movement:** WASD movement relative to the view, Space to jump, and gravity with collision against blocks. Mouse look uses pointer lock, and Esc releases the mouse.\n- **Interaction:** left click breaks the targeted block. Right click places the selected block on the targeted face, but not inside the player's body. A wireframe outline marks the target, and a DDA voxel raycast finds it within 6 blocks.\n- **Selection:** keys 1\u20133 select a block type. I also added the mouse wheel to cycle through them.\n- **UI:** a crosshair, a bottom hotbar that highlights the selected block, and a \"Selected: \u2026\" label.\n\nChoices I made beyond the spec:\n\n- Mouse look is implemented directly rather than with `PointerLockControls`, because that class lives in `examples/jsm` and would need a second import.\n- Shading is baked into vertex colours with `MeshBasicMaterial`, so the scene has no lights.\n- The mesh is rebuilt in full after each block edit. This is simple and fast enough for this world size.",
 "ttft_ms": 72603,
 "type": "result",
 "duration_ms": 102656,
 "uuid": "[REDACTED]",
 "ttft_stream_ms": 1140,
 "time_to_request_ms": 213,
 "first_content_frame_ms": 1140,
 "queued_turn_count": 0,
 "result_index": 0
}