Jevman Benchmark Tests Low-Latency Decision Models in Pac-Man
October 8, 2026
Jevman evaluates the real-time decision-making capabilities of models including GPT-6 Luna, Clef, and Jev 1.13 using Pac-Man. The benchmark highlights how low-latency architectures enable models to perform fast, reactive interactions in simulated environments.
HOW THIS AFFECTS YOU
●
builderThis demonstrates the utility of low-latency decision endpoints for real-time agentic tasks.
●
researcherThe leaderboard provides a concrete way to benchmark temporal reasoning and reactive decision-making in lightweight models.