Trajectory

1Add a light-beam puzzle chamber

ModelRun 1
Claude Sonnet 5.5Fail
GPT-6.1 SolFail
Claude Opus 5.5Fail
Gemini 4 ArgonFail
GPT-6 AstraFail
Claude Haiku 5.5Fail
Qwen3.8 MaxFail
Grok 4.7Fail

2Turn a blockout valley into a real place

ModelRun 1
Claude Sonnet 5.5Fail
GPT-6.1 SolFail
Claude Opus 5.5Fail
Gemini 4 ArgonFail
GPT-6 AstraFail
Claude Haiku 5.5Fail
Qwen3.8 MaxFail
Grok 4.7Fail

3A HUD, levelling, new enemy looks and a first village

ModelRun 1
Claude Sonnet 5.5Fail
GPT-6.1 SolPass
Claude Opus 5.5Fail
Gemini 4 ArgonPass
GPT-6 AstraPass
Claude Haiku 5.5Fail
Qwen3.8 MaxFail
Grok 4.7Fail

4The hero's fight, three levels, every menu and a hub village

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Pass
Gemini 4 ArgonFail
GPT-6 AstraFail
Claude Haiku 5.5Fail
Qwen3.8 MaxFail
Grok 4.7Fail

5Three levels, every menu and a hub village

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Pass
Gemini 4 ArgonPass
GPT-6 AstraFail
Claude Haiku 5.5Fail
Qwen3.8 MaxFail
Grok 4.7Fail

6Drift boosts, items, hazards, sound and menus for a buggy racer

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Fail
Gemini 4 ArgonPass
GPT-6 AstraPass
Claude Haiku 5.5Pass
Qwen3.8 MaxFail
Grok 4.7Fail

7Give the hero her fight

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Pass
Gemini 4 ArgonPass
GPT-6 AstraPass
Claude Haiku 5.5Pass
Qwen3.8 MaxFail
Grok 4.7Fail

8Playtest fixes for a fantasy battle game's first trial

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolFail
Claude Opus 5.5Fail
Gemini 4 ArgonFail
GPT-6 AstraFail
Claude Haiku 5.5Fail
Qwen3.8 MaxFail
Grok 4.7Fail

9Sound, feedback and guards for a mission played unseen

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolFail
Claude Opus 5.5Pass
Gemini 4 ArgonPass
GPT-6 AstraFail
Claude Haiku 5.5Pass
Qwen3.8 MaxPass
Grok 4.7Pass

10Playtest fixes for a helicopter demolition game

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Pass
Gemini 4 ArgonPass
GPT-6 AstraPass
Claude Haiku 5.5Pass
Qwen3.8 MaxPass
Grok 4.7Pass

11Play an action RPG against its spec and fix what breaks

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Pass
Gemini 4 ArgonFail
GPT-6 AstraPass
Claude Haiku 5.5Fail
Qwen3.8 MaxFail
Grok 4.7Fail

12Fix sixteen bugs from a QA pass

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Pass
Gemini 4 ArgonFail
GPT-6 AstraPass
Claude Haiku 5.5Pass
Qwen3.8 MaxPass
Grok 4.7Pass

13Fix the first look at a helicopter demolition game

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Pass
Gemini 4 ArgonPass
GPT-6 AstraPass
Claude Haiku 5.5Pass
Qwen3.8 MaxPass
Grok 4.7Fail

14Drop an event and act on playtest notes in an athletics game

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Pass
Gemini 4 ArgonPass
GPT-6 AstraPass
Claude Haiku 5.5Pass
Qwen3.8 MaxFail
Grok 4.7Fail

15Play an athletics game against its spec and fix it

ModelRun 1
Claude Sonnet 5.5Pass
GPT-6.1 SolPass
Claude Opus 5.5Pass
Gemini 4 ArgonPass
GPT-6 AstraPass
Claude Haiku 5.5Pass
Qwen3.8 MaxPass
Grok 4.7Fail