GLM 5.3
A large language model in Zhipu AI's GLM series, with a total parameter count of 320B and 18B activated parameters. It is positioned as a flagship model product within the GLM family.
Core Facts
Timeline (last 90 days)
On Artificial Analysis's new coding index, ML4 scores the same as GLM 5.3, ahead of other Western open-source models.
ML4 has a built-in length penalty mechanism and is 50% more efficient than GLM 5.3 on the coding index.
When adding models to a Combo, the selectable models include Gemini 2.5, Gemini 3 Flash, Claude Sonnet 4.5, Big Pickle, DeepSeek, GLM 5.3 and others.
DeepSeek V4.1 Flash scored 42 on Humanities Last Exam, tied with Kimi K3 and GLM 5.3; its Terminal Bench score was also 42.
Zhipu credits are converted based on Tokens, and the flagship model GLM 5.3 is about 3 times more expensive than GLM 5.3 Flash.
The active expert data volume of GLM 5.3 is 8.8GB, smaller than the 13.2GB of MiniMax M3.1 Flash.
On the uploader's RTX Pro 5000 platform, the speed performance of each model on Xtier is: Qwen 3.8 is the fastest with essentially no waste; DeepSeek 5.3 is slightly slower by 2% than the revised LS济浪; GLM 5.3 is faster than LS济浪; MiniMax M3.1 Flash is very slow.
OpenScience announced that it has exited beta and launched on Product Hunt; the new version brings a faster research workspace IDE and a model aggregation service covering more than 30 models, including GPT-6 Astra, Pimi K3, GLM 5.3, and others, all sharing a pay-as-you-go wallet; Auto Research can autonomously iterate experiments once a goal is given. The platform has been adopted by more than 30 universities and is free and open source.
GLM 5.3 not only showed screen flickering in this test, but the train also exhibited lateral movement and clipping, performing very poorly.
The researchers ran this experiment with GLM 5.3, and mentioned that OpenAI's and Anthropic's models as well as open-source models such as Kimi also performed well on similar tasks.
40 more timeline events
All Facts (20)
@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%VerifiedV4.1-Flash scores 1 point higher than a GPT-5.6-level model in comparison.
65%UnverifiedOn Artificial Analysis's new coding index, ML4 scores the same as GLM 5.3, ahead of other Western open-source models.
50%UnverifiedML4 has a built-in length penalty mechanism and is 50% more efficient than GLM 5.3 on the coding index.
50%UnverifiedWhen adding models to a Combo, the selectable models include Gemini 2.5, Gemini 3 Flash, Claude Sonnet 4.5, Big Pickle, DeepSeek, GLM 5.3 and others.
50%UnverifiedDeepSeek V4.1 Flash scored 42 on Humanities Last Exam, tied with Kimi K3 and GLM 5.3; its Terminal Bench score was also 42.
50%UnverifiedZhipu credits are converted based on Tokens, and the flagship model GLM 5.3 is about 3 times more expensive than GLM 5.3 Flash.
50%UnverifiedThe active expert data volume of GLM 5.3 is 8.8GB, smaller than the 13.2GB of MiniMax M3.1 Flash.
50%UnverifiedOn the uploader's RTX Pro 5000 platform, the speed performance of each model on Xtier is: Qwen 3.8 is the fastest with essentially no waste; DeepSeek 5.3 is slightly slower by 2% than the revised LS济浪; GLM 5.3 is faster than LS济浪; MiniMax M3.1 Flash is very slow.
50%UnverifiedGLM 5.3 not only showed screen flickering in this test, but the train also exhibited lateral movement and clipping, performing very poorly.
50%UnverifiedThe researchers ran this experiment with GLM 5.3, and mentioned that OpenAI's and Anthropic's models as well as open-source models such as Kimi also performed well on similar tasks.
50%UnverifiedZhipu's GLM Coding Plan is not an editor but a model plan, and one plan can connect to more than 20 coding tools such as Claude Code, Codex and Cursor; Lite is 118 yuan per month (94.4 yuan after a 20% annual discount), and Pro and Max have 6 times and 14 times the usage of Lite respectively; the model is GLM 5.3.
50%UnverifiedThe models currently available on Token Pass include DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3.
50%UnverifiedKimi K3 is provided by Moonshot AI, GLM 5.3 is provided by Z.ai, and DeepSeek V4.1 Flash is provided by DeepSeek.ai.
50%UnverifiedRunInfra does not train models itself; instead it integrates open-source models such as GLM 5.3, DeepSeek V4.1 Flash and Qwen3.8 27B.
50%UnverifiedPro subscribers of Vibe Coder Planner can directly use the built-in GLM 5.3 development model without separately applying for an API key or configuring accounts with other AI providers.
50%UnverifiedAccording to the information published in the original article, Vibe Coder Planner Pro was priced at $4.99 per month at the time, including unlimited project planning during the subscription period, all task prompts, and the GLM 5.3 model.
50%UnverifiedScaleway's Generative APIs offer more than 40 available models, with LightOn OCR 3, DeepSeek V4.1-Flash and GLM 5.3 newly added this time.
50%UnverifiedAccording to the creator's real-world tests and community feedback, DeepSeek-V4.1-flash has surpassed Claude Opus 5, GPT 5.6 Sol, and a range of domestic open-source models on core metrics measuring real-world programming capability.
50%UnverifiedMuse Spark Open scored 76.4 on Terminal-Bench 2.1, lower than GLM 5.3's 83.9 and Kimi K3's 85.
50%