AI BriefingStarCraft AI Benchmark 'Brood War Bench' Released... Only Codex Astra xhigh Achieves 18-0 Record
Key point
In the Brood War Bench evaluating AI agents' StarCraft performance, the Codex Astra xhigh model took first place with 18 wins in 18 matches, while Claude Haiku and Grok 4.6 low remained at the bottom with 0 wins.
Details
Results from 'Brood War Bench', which evaluates AI agents' StarCraft gameplay performance, have been released. The Codex Astra xhigh model secured first place with 18 wins in 18 matches (100% win rate), establishing the only undefeated record. This model recorded an average APM of 12.6 and a cost of $10.54 per game. Claude Fable came in second with an 83.3% win rate, while the Grok 4.6 series showed low performance with win rates of 11.1% or less. Notably, Claude Haiku and Grok 4.6 low recorded 16 losses in 16 matches (0 wins). Detailed metrics analysis showed that Codex Astra maintained an average tech investment of 0.3, 14.7 workers, and 8.3 army units, whereas Grok 4.6 managed only 0 tech investment, 7.4 workers, and 2 army units. Additionally, Grok 4.6 lost with an average of 536.6 unused resources, while Codex Astra demonstrated relatively higher resource utilization efficiency with 240.6.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
SWE-Bench Pro V2 Released: 89 Tasks Removed and Evaluation Protocol StrengthenedQwen Releases 'RecreationWorld' for Training AI Agents to Rebuild Apps
US Appeals Court Upholds Pentagon Blacklisting of Anthropic as Supply Chain Risk