# zai-org/GLM-4

GLM-4 series: Open Multilingual Multimodal Chat LMs | 开源多语言多模态对话模型

Repository: https://github.com/zai-org/GLM-4
Canonical: https://ross.abutalabs.com/products/glm-4
Language: Python
License: Apache-2.0
License Family: permissive
Topics: chatglm, chatglm-6b, glm, glm-4, glm4
Last push: 2026-08-05T07:08:46+00:00

## Health v2 (maintenance only)
Score: 67/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 96, release rhythm 35, longevity 60
- inputs: {"age_days": 840, "days_push": 28, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 7070, forks 611 (observed 2026-08-28T04:09:55.682784+00:00)

## What it is
Official repository for the GLM-4 series of open-weight, multilingual (primarily Chinese/English) multimodal chat language models from Z.ai (Zhipu AI), spanning base chat models like GLM-4-9B and GLM-4-32B, the GLM-Z1 reasoning variants, and GLM-4V vision-language models. It ships model weights alongside inference examples, fine-tuning scripts, and deployment guides for frameworks like transformers and vLLM.

## Use cases
- run a GLM-4 chat model locally on my own GPU
- self-host an open-source alternative to GPT-4o
- fine-tune a multilingual chat LLM for my domain
- download GLM-4 or GLM-Z1 model weights from Hugging Face
- deploy a bilingual Chinese-English dialogue model
- use a reasoning LLM for math, code, and logic tasks
- set up an LLM with function calling for agent workflows

## When to choose
- You want open-weights chat LLMs you can deploy on-premises or on your own hardware
- You need strong bilingual Chinese/English dialogue, reasoning, or coding capabilities at 9B-32B scale
- You want to fine-tune or apply reinforcement learning to a mid-size agent-capable model
- You need vision-language (GLM-4V) or deep-thinking (GLM-Z1) variants from the same model family

## When to avoid
- You only want a hosted commercial API without managing GPUs or model weights
- You require frontier-scale performance beyond what 32B open weights deliver
- Your hardware cannot accommodate 9B+ parameter models even in quantized form
- You need a fully managed, multimodal production service rather than DIY deployment

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, llm-training, chatbot, deep-learning, nlp, agent-framework
- domain: artificial-intelligence, large-language-models, machine-learning, deep-learning, chatbots
- platform: python
- tags: open-weights, glm-4, chatglm, multilingual, multimodal, fine-tuning, transformers, vllm, reasoning-model, function-calling, self-hosted-llm, huggingface, zhipu-ai, natural-language-processing, gpu

## Member repositories
- zai-org/GLM-4 (main) score 67

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:55.682784+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:40:27.729798+00:00, confidence not recorded.
  - readme: https://github.com/zai-org/GLM-4 (fetched 2026-08-28T04:09:55.682784+00:00, sha 547a67201178)
- Data as of 2026-08-30T08:39:29.467469+00:00.