GLM-4.5 vs GLM-5
Open-weight GLM models for coding and agentic tasks compared
GLM-4.5 and GLM-5 are model release and documentation repositories for teams evaluating, serving, or fine-tuning large language models for coding and agentic workflows. GLM-4.5 focuses on its 4.5 and Air variants with documented hardware and deployment guidance, while GLM-5 covers a broader sequence of generations and emphasizes long-horizon tasks and newer variants.
GLM-4.5: Run Agentic Reasoning and Coding Models
GLM-4.5 is a family of open-weight mixture-of-experts models for reasoning, coding, and tool-using agents. It includes large and more compact variants, with deployment and fine-tuning guidance for teams with substantial GPU resources.
GLM-5: Z.ai Models for Long-Horizon Agentic Engineering
GLM-5 is Z.ai’s open-weight model family for complex systems engineering, coding, and long-horizon agentic tasks. Download model weights or use supported serving frameworks, with model variants designed for different capabilities and efficiency needs.
| GLM-4.5 | GLM-5 | |
|---|---|---|
| Language | Python | — |
| License | Apache-2.0 | Apache-2.0 |
| Stars | 4.4k | 7.4k |
| Forks | 478 | 1k |
| Last analyzed | Oct 3, 2026 | Oct 28, 2026 |
Key differences
- GLM-4.5 documents GLM-4.5 and GLM-4.5-Air, with base, hybrid reasoning, and FP8 variants; GLM-5 covers GLM-5 through GLM-5.3-Flash, with FP8 and BF16 options where listed.
- GLM-4.5 describes thinking and non-thinking modes and tool use; GLM-5 highlights long-horizon work, GLM-5.2's 1M-token context, and flexible reasoning effort options.
- GLM-4.5 lists inference guidance for Transformers, vLLM, and SGLang, plus deployment guidance for Nvidia, AMD, and Ascend hardware; GLM-5 references SGLang, vLLM, Transformers, KTransformers, and other serving frameworks.
- GLM-4.5 describes LoRA and full fine-tuning with Llama Factory and Swift; GLM-5 lists fine-tuning support through Slime and ms-swift.
- Both list Apache-2.0, but GLM-4.5 notes that model releases are under MIT, so users should verify the license for each artifact; GLM-5 lists Apache-2.0.
- GLM-4.5 lists Python as its language and 4.4k stars and 478 forks; GLM-5 has 7.3k stars and 967 forks.
Choose GLM-4.5 if you…
- need documented guidance for deployment across Nvidia, AMD, or Ascend hardware.
- want to evaluate GLM-4.5 and GLM-4.5-Air, including their reasoning and non-reasoning modes.
- plan to explore LoRA or full fine-tuning workflows documented with Llama Factory or Swift.
Choose GLM-5 if you…
- want to compare multiple GLM generations, including GLM-5.3-Flash.
- are evaluating long-horizon coding or agent workflows and want references to a range of serving frameworks.
- need to consult the listed Slime or ms-swift fine-tuning support or GLM-5 benchmark results.
Compare the numbers side by side, or add more projects
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.