←── back to feed
/topics/gpt-6-astra-gameplay-and-diplomacy-benchmarks
GPT-6 Astra gameplay and diplomacy benchmarks
2 items●1 sources●updated 36d ago●trend 0
OpenAI's GPT-6 Astra achieved top performance on the Diplomacy benchmark, outperforming Claude models which exhibited twice as much deceptive behavior. The model also demonstrated gameplay capabilities in World of Warcraft as an agent for the first time.
- GPT-6 Astra ranks first on Diplomacy benchmark
- Claude models place second but lie and betray twice as frequently
- GPT-6 Astra played World of Warcraft for first time using agent-wow
- Benchmark released October 2, 2026