GLM
GLM (General Language Model) is a large language model series developed by Zhipu AI, built on an autoregressive blank-infilling pre-training framework. The series supports natural language understanding, text generation, conversational interaction, and code writing, with bilingual Chinese-English processing capability. The GLM series encompasses versions of varying parameter scales and can be applied to question answering, summarization, reasoning, and other natural language processing tasks.
Core Facts
@ai-sdk/zai is the official Provider in the Vercel AI SDK for Zhipu AI models.
GLM unifies the pre-training objectives of natural language understanding and generation through Autoregressive Blank Infilling.
The GLM series originally adopted an autoregressive blank-filling pre-training framework that accommodates both understanding and generation capabilities.
Timeline (last 90 days)
Z.ai is an international product line (overseas brand) launched in 2024 by Zhipu AI (the Chinese large-model company Zhipu), and its GLM series models perform close to the international frontier in multiple benchmarks.
Over the past year or more, open-source large models from China have frequently appeared on the Hugging Face trending list, including DeepSeek, Qwen, GLM, MiniCPM and others.
GLM-4.7-Flash scores about 59% on SWE-bench and can run on a 5-8GB laptop.
SEO Master Agent supports OpenAI, Anthropic, Gemini, GLM, and Ollama simultaneously through a multi-LLM gateway, with FastAPI on the backend, Next.js on the frontend, and asynchronous task execution.
The Tongyi Qianwen Flash edition starts at $0.03 and $0.13, and GLM scored 1,647 points on GDPVal.
The author summarizes the personalities of models from each vendor: GLM is pessimistic and overly logical, Qwen is overly inquisitive and flashy, Gemma is witty and creative, Mistral is flirtatious, and Llama is formulaic and predictable.
Zhipu plans to scale its next-generation GLM model from approximately 740 billion parameters to 1 trillion parameters and to build a closed-loop recursive self-improvement system.
Domestic or open-source models such as GLM and DeepSeek have fewer restrictions on tasks like vulnerability analysis, partly due to differences in training data and alignment strategies, but may be more permissive in other safety boundaries.
Reflection stated that Beam achieves comparable scores to GLM on reasoning benchmarks while consuming three to four times less inference compute, a figure estimated based on activated parameters and generated token counts rather than measured costs.
The poster believes that if recurrent architectures can work stably at industrial scale, open-source model teams such as DeepSeek, Qwen, and GLM may begin experimenting with this type of architecture.
40 more timeline events
All Facts (20)
@ai-sdk/zai is the official Provider in the Vercel AI SDK for Zhipu AI models.
90%VerifiedGLM unifies the pre-training objectives of natural language understanding and generation through Autoregressive Blank Infilling.
90%VerifiedThe GLM series originally adopted an autoregressive blank-filling pre-training framework that accommodates both understanding and generation capabilities.
90%VerifiedEarly GLM adopted Autoregressive Blank Infilling as its pre-training objective
90%VerifiedGLM系列由清华大学KEG实验室与智谱AI联合研发
90%VerifiedThe GLM series is a large language model jointly developed by Tsinghua University and Zhipu AI, built on the Autoregressive Blank Infilling training objective.
90%Verified@ai-sdk/zai depends on @ai-sdk/openai-compatible, indicating that Zhipu GLM's API interface is largely compatible with OpenAI's calling conventions.
85%VerifiedVLMs (Vision Language Models) typically consist of a visual encoder (such as ViT), a language model, and a projection layer connecting the two, with representative models including GPT-4V/4o, Claude 3.5, the LLaVA series, and InternVL
80%VerifiedGLM采用了独特的自回归填空预训练范式，与GPT的纯自回归和BERT的纯掩码语言模型都有所不同
80%VerifiedThe GLM series models from Zhipu (Z.ai) were jointly developed by Tsinghua University and Zhipu AI
75%VerifiedZhipu AI is an artificial intelligence company established through the commercialization of technology from Tsinghua University's KEG Lab.
75%VerifiedGLM uses an autoregressive blank-filling training objective, randomly masking text segments and requiring the model to autoregressively restore them.
75%VerifiedZhipu AI is a technology commercialization spinoff of Tsinghua University, and its GLM architecture adopts an autoregressive blank-filling pre-training objective.
75%VerifiedDeepSeek Harness supports connecting to third-party providers, allowing access to models such as GLM via OpenCode Go, or connection to a local Ollama service.
65%VerifiedChatGLM is developed by Zhipu AI and is based on the GLM architecture.
65%VerifiedZhipu's GLM series models (including GLM-4 and ChatGLM) are large language models jointly developed by Tsinghua University and Zhipu AI.
65%Unverified智谱AI的GLM系列脱胎于清华大学计算机系的学术研究
90%UnverifiedDomestic models such as GLM have increased the proportion of high-quality Chinese corpus during the training phase and optimized the tokenizer design for Chinese characters.
70%UnverifiedZhipu AI is an artificial intelligence company incubated by a technical team from Tsinghua University; its GLM series of large models is based on a self-developed GLM pre-training architecture and uses an autoregressive blank-filling training approach.
70%UnverifiedZCode is a programming Agent based on the GLM model, focused on using large models to assist with code understanding, generation, and refactoring, and is capable of reading project context to provide intelligent suggestions.
60%