According to 1M AI News monitoring, today the global authoritative AI evaluation platform LMArena (with over one million users participating in blind tests) updated its Code Arena special ranking, with GLM-5.1 topping the global open-source model list and ranking third among all models worldwide.
GLM-5.1 not only inherits the previous generation model’s open-source SOTA coding capabilities but also makes breakthroughs in long-horizon tasks, achieving:
Building a Linux desktop from scratch in 8 hours;
655 iterations breaking through the optimization bottleneck of vector databases;
1000 rounds of tool invocation optimizing real machine learning model loads.
It is worth mentioning that under the same evaluation standards on the METR leaderboard, GLM-5.1 is the only open-source model capable of sustained work for 8 hours, and is one of the few models in the world besides Claude Opus 4.6 that possesses this capability.
GLM-5.1がLMArenaコードランキングでオープンソース第1位、世界第3位
According to 1M AI News monitoring, today the global authoritative AI evaluation platform LMArena (with over one million users participating in blind tests) updated its Code Arena special ranking, with GLM-5.1 topping the global open-source model list and ranking third among all models worldwide.
GLM-5.1 not only inherits the previous generation model’s open-source SOTA coding capabilities but also makes breakthroughs in long-horizon tasks, achieving:
It is worth mentioning that under the same evaluation standards on the METR leaderboard, GLM-5.1 is the only open-source model capable of sustained work for 8 hours, and is one of the few models in the world besides Claude Opus 4.6 that possesses this capability.