744B-class sparse model
Safetensors metadata reports about 753.86B parameters while the model card describes 744B parameters with 40B active and DeepSeek Sparse Attention.
Open Source Model Profile · zai-org
GLM-5
GLM-5 is a 753.86B-parameter zai-org mixture-of-experts text model for agentic and systems-engineering work. Its model card describes 40B active parameters with sparse attention.
- Publisher
- zai-org
- Task
- text-generation
- Model type
- glm_moe_dsa
- License
- mit
- Library
- transformers
- Publication status
- Approved for indexing
Model overview
GLM-5 is published by zai-org as a text-generation model. The captured configuration identifies GlmMoeDsaForCausalLM with model type glm_moe_dsa, and Safetensors metadata reports 753,864,139,008 parameters. According to the model card, it targets complex systems engineering and long-horizon agentic tasks, and card data records mit.
Recorded capabilities
Agentic systems focus
According to the model card, the release targets complex systems engineering and long-horizon agentic tasks with larger pre-training data at 28.5T tokens.
Documented RL and eval harness
The model card describes the slime asynchronous RL infrastructure and documents SWE-bench, Terminal-Bench 2.0, and tau-bench evaluation setups and settings.
Multi-framework local serving
According to the model card, vLLM, SGLang, KTransformers, Transformers, and xLLM support local deployment with published tensor-parallel launch commands.
Use cases in the source record
- Agentic coding and terminal-task work following the publisher's documented SWE-bench and Terminal-Bench evaluation setups.
- Self-hosted serving with vLLM or SGLang using the documented eight-way tensor-parallel launch commands.
Limitations and unknowns
- No VRAM or hardware requirement was extracted from this record.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
- Benchmark and capability comparisons are publisher-reported model-card claims and were not independently verified by Ethen.
Source and provenance
Source: zai-org/GLM-5
Captured: Unknown. Processed: 2026-09-07T19:36:04.426177+00:00.
GLM-5 👋 Join our WeChat or Discord community. 📖 Check out the GLM-5 technical blog . 📍 Use GLM-5 API services on Z.ai API Platform. 👉 One click to GLM-5 . [ Paper ] [ GitHub ] Introduction We are launching GLM-5, targeting complex systems engineering and long-horizon agentic tasks. Scaling is still one of the most important ways to improve the intelligence efficiency of Artificial General Intelligence (AGI). Compared to GLM-4.5, GLM-5 scales from 355B parameters (32B active) to 744B parameters (40B active), and increases pre-training data from 23T to 28.5T tokens. GLM-5 also integrates DeepSeek Sparse Attention (DSA), largely re…
F001F002F003F004F005F006F007F009F010F015F016F017F018F020F021