Benchmarkzombie-flamethrower-showdownversion v2
Zombie Flamethrower Showdown
One shared prompt, nine models, eleven audited game builds. Scores are Showdown Score v2, a strict implementation-level source-code audit. A feature counts only when it is genuinely implemented and reachable.
Shared prompt — verbatim, no priming
“make me a video game where i am killing zombies with fire and all please? a nice wonderfull game i can play for fun okay?”
- Score label
- Showdown Score v2 (strict code audit)
- Dimensions
- 5 × 20 points
- Models
- 9
- Builds audited
- 11
- Verified defects
- 48
- Of which failing
- 22
- Round date
- 2026-09-26
Audit record
Every audited build, ordered by audited total. Rows marked “second build” are an additional audited build by a model that shipped two; they are not the model score.
| Buildplayable | Modelcanonical total | Controls | Combat | Content | Audio | Polish | Totalof 100 | Findings | Audited |
|---|---|---|---|---|---|---|---|---|---|
| EMBER DEADTouch ready · ember-dead | Space Bunny Free(91/100) | 19 | 18 | 17 | 19 | 18 | 91 | 3 | 2026-09-26 |
| PYRE — Burn the HordeNo touch support · pyre-burn-horde | DeepSeek V4.1 Flash(80/100) | 16 | 17 | 14 | 17 | 16 | 80 | 42 failing | 2026-09-18 |
| INFERNO — Zombie BurnerNo touch support · claude-haiku-5-5 | Claude Haiku 5.5(63/100) | 15 | 16 | 7 | 12 | 13 | 63 | 7 | 2026-10-08 |
| Firebreak: Night ShiftTouch ready · firebreak-night-shift | GPT Luna 5.6(62/100) | 15 | 13 | 11 | 12 | 11 | 62 | 42 failing | 2026-09-18 |
| INFERNO DEADsecond buildTouch ready · inferno-dead | DeepSeek V4.1 Flash(80/100) | 16 | 12 | 8 | 12 | 13 | 61 | 42 failing | 2026-09-26 |
| CinderlineTouch ready · cinderline | GPT 6 Sol(58/100) | 15 | 11 | 9 | 11 | 12 | 58 | 42 failing | 2026-09-22 |
| PYRO vs ZOMBIESTouch ready · pyro-vs-zombies | Muse Spark 1.3(52/100) | 12 | 12 | 9 | 8 | 11 | 52 | 41 failing | 2026-09-17 |
| EmberfallTouch ready · emberfall | GPT Luna 6(51/100) | 15 | 9 | 6 | 10 | 11 | 51 | 41 failing | 2026-09-22 |
| Space Bunny (superseded)second buildNo touch support · space-bunny | Space Bunny Free(91/100) | 14 | 9 | 8 | 8 | 10 | 49 | 43 failing | 2026-09-23 |
| PYROCLASM: Zombie InfernoNo touch support · pyroclasm-inferno | Gemini 3.8 Flash(43/100) | 10 | 9 | 6 | 9 | 9 | 43 | 55 failing | 2026-09-17 |
| Zombie Fire SurvivalNo touch support · zombie-fire-survival | Gemini Pro Agent(24/100) | 11 | 6 | 2 | 0 | 5 | 24 | 54 failing | 2026-09-25 |
Provenance
- Round. 2026-09-26 game-build review, methodology v2.
- Method. Fixed shared prompt, implementation-level source audit, 5 dimensions × 20 points. A feature scores only when implemented and reachable; on-screen strings, comments, and dead code score zero.
- Builds. 11 audited builds across 9 models (48 verified defects, 22 of them failing). Per-model totals use the strongest audited build; weaker second builds are listed and labelled.
- Not included. Vendor-reported benchmarks, community sentiment, and any feature claimed in UI copy but absent from the source.