Results / Qwen3-4B Q4_0 (Pi CPU, 2 A76 threads)
Qwen3-4B Q4_0 (Pi CPU, 2 A76 threads)
pi-builderllama.cppOrange Pi 5 Pluspi+steer-v2
The game opens in a new tab. It runs sandboxed, so a game that saves to browser storage may not keep its save.
Metrics
- Checklist score
- – / 100
- Design (rubric, 0–5)
- –
- Stop reason
- stuck
- Model time
- 6hr 5mins
- Tool time (wall minus model)
- 14hr 14mins
- Wall time
- 20hr 19mins
- Tool calls
- 51
- Tool errors
- 37
- Errors by type
- {"guard": 31, "other": 5, "format": 1}
- Loop-guard events
- 30
- Tokens in / out
- 13,758 / 8,826
- Thinking tokens
- 0
- Decode / prefill tok/s
- 0.6 / 2.2
- Prefix reuse
- –
- Peak VRAM (GPU / engine)
- – / – MiB
- Peak engine RAM
- – MiB
- Engine
- llama.cpp
- Model
- Qwen3-4B Q4_0 (Pi CPU, 2 A76 threads) (qwen3-4b-q4_0)
- Settings
- {"ctx": 8192, "step_limit": 30, "max_tokens": 4096, "hint_after": 4, "stop_after": null}
- Node
- Orange Pi 5 Plus
- Prompt
- v1
- Started
- 2026-10-07T10:57:23Z
Feature scores
Screenshots
Landing page at desktop and phone widths, and the game at noon, at night, after breaking a block, and on a phone.
Checklist (0 pass, 0 partial, 0 fail)
| Feature | Test | Score | Evidence |
|---|
Suite · · WebGL:
Steers
None: an unsteered run gets the prompt and nothing else.