GLM models

Z.ai's GLM models: GLM-5.3 is a leading open coding model, GLM-5.3-Flash is multimodal, and GLM-4.7-Flash runs on a single 24 GB GPU.

Open weights
Models 5
Context up to 1M
Updated Nov 2026

Hugging Face Compare

On this page

At a glance
Models & versions
VRAM & download size
Run it locally
Best uses
Strengths & limits
About GLM
Guides
FAQ
Compare

At a glance

Type: Open-weight models
Developer: Z.ai (Zhipu AI) (China)
First release: Mar 2023
Models tracked: 5 (4 current)
Longest context: 1M tokens
Abilities: Tool callingReasoning modeImage input
Licence: MIT for GLM-5.2, GLM-5.3-Flash, GLM-4.7-Flash and GLM-OCR. GLM-5.3 uses its own GLM-5.3 licence.
Smallest local option: GLM-OCR Default · ~2.5 GB
Website: z.ai
Hugging Face: huggingface.co/zai-org

Newest model

GLM-5.3

Released Aug 2026

Smallest option

GLM-OCR Default

~2.5 GB VRAM · 1.8 GB download

GLM models and versions

Model	Released	Sizes	Context	Abilities	Licence	Status
GLM-5.3 Flagship open coding model with 1M context	Aug 2026	753B Too large for most servers · Ollama cloud	—	ToolsReasoning	GLM-5.3 License	Current
GLM-5.3-Flash First multimodal GLM-5 model; strong coding at low cost	Aug 2026	320B 320B total, 18B active (MoE) Too large for most servers · Ollama cloud	—	ToolsReasoningVision	MIT	Current
GLM-4.7-Flash Strong 30B-class model for one GPU	Jan 2026	30B 30B total, 3B active (MoE)	198K	ToolsReasoning	MIT	Current
GLM-OCR Document OCR and layout understanding	Jan 2026	Default	128K	ToolsVision	MIT	Current
GLM-5.2 Long-horizon tasks with a stable 1M context	Jun 2026	Too large for most servers · Ollama cloud	1M	ToolsReasoning	MIT	Previous generation

Sizes, context windows and abilities come from the Ollama library and Hugging Face model cards, checked 14 Nov 2026. Release months are the first public upload. Spotted a mistake? Tell us.

How much VRAM does GLM need?

Download sizes per Ollama build and the GPU memory each needs with an 8K context. On a CPU-only server the same number is the RAM you need. Longer contexts need more memory; work it out with the LLM VRAM calculator.

Model	4-bit (default)	8-bit	16-bit	VRAM (4-bit)	Fits on	Command
GLM-4.7-Flash	19 GB	32 GB	60 GB	~21.5 GB	24 GB GPU
GLM-OCR	1.8 GB	1.8 GB	2.7 GB	~2.5 GB	8 GB GPU

VRAM = 4-bit download + about 10% runtime overhead + 0.5 GB for an 8K context, rounded up. Mixture-of-experts (MoE) models still need all their weights in memory, but they run much faster than a dense model of the same size, so they also work well split between GPU and system RAM.

Run GLM-4.7-Flash locally

# Install Ollama (Linux), then pull and chat
curl -fsSL https://ollama.com/install.sh | sh
ollama run glm-4.7-flash

New to Ollama? Follow Install Ollama on Ubuntu 26.04 (or 24.04) for GPU drivers, the systemd service and remote access.

# Ollama serves an OpenAI-compatible API on port 11434
curl http://localhost:11434/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "glm-4.7-flash", "messages": [{"role": "user", "content": "Hello"}]}'

# Full-precision weights from Hugging Face, OpenAI-compatible API on :8000
vllm serve zai-org/GLM-4.7-Flash

vLLM loads the original 16-bit weights, so it needs roughly the 16-bit size above in GPU memory, plus room for the context. See Install vLLM on Ubuntu 26.04.

What GLM is good for

Coding agents

GLM-5.3 leads open models on terminal and SWE agent benchmarks according to Z.ai.

Local agent on 24 GB

GLM-4.7-Flash (3B active) is fast and capable for tool calling.

Document processing

GLM-OCR is a small model for reading complex documents.

Strengths and limits

Strengths

MIT licence on most models.
1M-token context on GLM-5.x.
Good balance of quality and cost on the Flash models.

Watch out for

GLM-5.x models need data-centre GPUs.
GLM-5.3 has a custom licence, unlike the rest of the family.

About GLM

GLM (“General Language Model”) comes from Z.ai, formerly Zhipu AI, a Beijing company spun out of Tsinghua University. Its ChatGLM models were early open bilingual models; the GLM-4.5 series in 2025 made it one of the top open-weight labs.

GLM-4.7-Flash (January 2026) is a 30B mixture-of-experts model with 3B active parameters that runs on a 24 GB GPU. GLM-5.2 (June 2026) is built for long-horizon work with a 1M-token context. GLM-5.3 (August 2026) uses the same base model with heavier post-training for coding, and GLM-5.3-Flash is the first natively multimodal GLM-5 model (320B total, 18B active), released under MIT.

Guides for running GLM

Tool LLM VRAM Calculator: How Much GPU Memory Do You Need? Updated Nov 2026 AI & LLMs How to Install Ollama on Ubuntu 26.04 and Run LLMs Locally Updated Nov 2026 AI & LLMs How to Install llama.cpp on Ubuntu 26.04 (CPU and NVIDIA GPU) Updated Nov 2026 AI & LLMs How to Install vLLM on Ubuntu 26.04 (OpenAI-Compatible LLM Server) Updated Nov 2026 AI & LLMs How to Install Open WebUI on Ubuntu 26.04 (Self-Hosted ChatGPT Alternative) Updated Nov 2026

Frequently asked questions

What is the latest GLM model?

The newest GLM model is GLM-5.3, released in Aug 2026.

How much VRAM do I need to run GLM locally?

The smallest option, GLM-OCR Default, is a 1.8 GB download at 4-bit and needs about 2.5 GB of GPU memory (or RAM on a CPU-only machine) with an 8K context. Larger sizes need more; see the hardware table on this page, and use the VRAM calculator for longer contexts.

How do I run GLM with Ollama?

Install Ollama, then run "ollama run glm-ocr". Ollama downloads the 4-bit build and starts a chat; the same model is also available on the local API at http://localhost:11434.

Can I use GLM commercially?

MIT for GLM-5.2, GLM-5.3-Flash, GLM-4.7-Flash and GLM-OCR. GLM-5.3 uses its own GLM-5.3 licence. Always read the licence of the exact model you deploy; it is linked in the models table.

Compare GLM

GLM vs Qwen

← All AI models