GLM-5, newly released as open source, signals a broader shift in artificial intelligence. Large language models are moving beyond generating code snippets or interface prototypes toward building complete systems and carrying out complex, end-to-end tasks. The change marks a transition from so-called “vibe coding” to what researchers increasingly describe as agentic engineering.
This press release features multimedia. View the full release here: https://www.businesswire.com/news/home/20260215030665/en/
LLM Performance Evaluation: Agentic, Reasoning and Coding
Built for this new phase, GLM-5 ranks among the strongest open-source models for coding and autonomous task execution. In practical programming settings, its performance approaches that of Claude Opus 4.5, particularly in complex system design and long-horizon tasks requiring sustained planning and execution.
The model rests on a new architecture aimed at scaling both capability and efficiency. Its parameter count has expanded from 355bn to 744bn, with active parameters rising from 32bn to 40bn, while pre-training data has grown to 28.5trn tokens. These increases are paired with advances in training methods. A framework called Slime enables asynchronous reinforcement learning at a larger scale, allowing the model to learn continuously from extended interactions and improve post-training efficiency. GLM-5 also introduces DeepSeek Sparse Attention, which maintains long-context performance while cutting deployment costs and improving token efficiency.
Benchmarks suggest strong gains. On SWE-bench-Verified and Terminal Bench 2.0, GLM-5 scores 77.8 and 56.2, respectively, the highest reported results for open-source models, surpassing Gemini 3 Pro in several software-engineering tasks. On Vending Bench 2, which simulates running a vending-machine business over a year, it finishes with a balance of $4,432, leading other open-source models in operational and economic management.
These results highlight the qualities required for agentic engineering: maintaining goals across long horizons, managing resources, and coordinating multi-step processes. As models increasingly assume these capabilities, the frontier of AI appears to be shifting from writing code to delivering functioning systems.
Chat & Official API Access
Z.ai Chat: https://chat.z.ai
GLM Coding Plan: https://z.ai/subscribe
Open-Source Repositories
GitHub: https://github.com/zai-org/GLM-5
Hugging Face: https://huggingface.co/zai-org/GLM-5
Blog
GLM-5 Technical Blog: https://z.ai/blog/glm-5
View source version on businesswire.com: https://www.businesswire.com/news/home/20260215030665/en/
© Business Wire, Inc.
Disclaimer :
هذا البيان الصحافي ليس وثيقة من إعداد وكالة فرانس برس. لن تتحمل وكالة فرانس برس أية مسؤولية تتعلق بمضمونه. ألرجاء التواصل مع الأشخاص/المؤسسات المذكورين في متن البيان الصحافي في حال كانت لديكم أية أسئلة عنه.