GLM-5.1 Ranked First in LMArena Code Innovation Ranking, Third Globally

Leading global AI benchmarking platform LMArena (which has over 1 million users participating in blind tests) updated its Code Arena specialized leaderboard today, per monitoring from 1M AI News. GLM-5.1 topped the global open-source model rankings and took third place overall in the global model rankings. GLM-5.1 not only retains the state-of-the-art (SOTA) coding capabilities of its predecessor in the open-source space but also made significant breakthroughs in long-horizon tasks, including: 1. Building a Linux desktop from scratch in 8 hours; 2. Overcoming the vector database optimization bottleneck in 655 iterations; 3. Optimizing real-world machine learning model workloads via 1,000 rounds of tool calls. Notably, under the same evaluation metrics as the METR ranking, GLM-5.1 is the only open-source model capable of sustained work at the 8-hour mark—and one of just a handful of models globally with this capability, alongside Claude Opus 4.6.