"Z.ai claims that GLM-5.1 can sustain performance over hundreds of iterations, distinguishing it from other models that lose effectiveness in longer sessions."
"The model improved a vector database optimization task over more than 600 iterations, achieving 21,500 queries per second, significantly surpassing previous results."
"GLM-5.1 outperformed its predecessor on several benchmarks, scoring 58.4 on SWE-Bench Pro, exceeding the scores of OpenAI's GPT-5.4 and others."
"Z.ai emphasizes that long-running performance is a key differentiator, as many current models plateau after a limited number of turns."
Z.ai introduced GLM-5.1, an open-source coding model aimed at agentic software engineering. This model can maintain performance over hundreds of iterations, unlike others that decline in effectiveness. GLM-5.1 excelled in software engineering benchmarks, outperforming its predecessor and competing models. It achieved significant results in tasks like vector database optimization, reaching 21,500 queries per second. Released under the MIT License, it offers enterprises more control over deployment, addressing the limitations of current models that plateau after a few turns.
Read at InfoWorld
Unable to calculate read time
Collection
[
|
...
]