StarSkirmish Bench has been refreshed with results from flagship LLMs (OpenAI, Anthropic, Google, Muse) playing StarCraft: BW by writing code. The benchmark evaluates how well AI models can implement real-time strategy game logic through code generation.
Background
StarSkirmish is a benchmark initiative where LLMs compete by generating code to play StarCraft: Brood War, testing reasoning, planning, and code execution capabilities in a real-time environment.
- Source
- Hacker News (RSS)
- Published
- Oct 16, 2026 at 10:18 PM
- Score
- 6.0 / 10