Zhipu AI chat: the web chat interface overview
A product-level reference for the Zhipu AI chat interface — covering its features, conversation history, model picker, export functionality, browser support, and accessibility profile.
5+
GLM variants
History
Saved per account
Export
MD & plain text
26+
Languages
The Zhipu AI chat product surface
Zhipu AI chat is the web chat interface for the Z.ai platform — distinct from the workflow patterns described on the chat Z AI page, this reference covers the interface itself.
The Zhipu AI chat interface is the primary product surface for users who want to interact with the GLM model family without any developer tooling. It is a full-featured web application rather than a minimal demo: signed-in accounts get persistent conversation history, a model picker that spans the current GLM variant lineup, a prompt recall system that surfaces recent inputs, and the ability to export individual conversations for archiving or sharing. Visitors who arrive without an account can start a conversation immediately, but without persistence or the full feature set.
The product relationship to the underlying platform is worth understanding. Zhipu AI chat is the consumer-facing product layer sitting on top of the same inference infrastructure that powers the Zhipuai API. The models are identical; what differs is the delivery mechanism, the session management, and the set of controls exposed. The Zhipu AI open platform is the developer-facing console for billing and key management; Zhipu AI chat is where end users and non-developers interact with the models directly.
Spotlight Brief
The Zhipu AI chat interface stores conversation history per signed-in account, not per browser session. Signing in before starting a conversation you want to keep is the single most important setup step. Conversations started as a guest cannot be recovered after the browser tab closes — they are not attached to any account and are not retrievable through the login flow.
Zhipu AI chat interface features
Five features define the Zhipu AI chat interface beyond basic text exchange — each is worth understanding before you rely on it for a workflow.
Conversation history
The Zhipu AI chat sidebar lists all conversations for the signed-in account in reverse chronological order. Each conversation is accessible with a single click and resumes exactly where it left off, including the full message thread and the model that was active at the last turn. The sidebar supports renaming conversations for organisation and deleting ones no longer needed. There is no hard cap on the number of stored conversations, though very old inactive threads may be subject to archival depending on account tier.
Model picker
The model picker in the Zhipu AI chat interface sits at the top of the conversation pane and lists the current GLM variant lineup. Switching models starts a new conversation; the prior conversation is not retroactively re-run with the new model. The picker shows the display name and a brief capability descriptor for each variant, which helps users choose between the general-purpose GLM-4.5+, the code-specialised variant, and the smaller faster variants suited for high-iteration drafting work.
Prompt history and recall
Zhipu AI chat surfaces a prompt history feature that lists recent inputs from the current account. This is separate from conversation history — it shows individual prompts rather than full conversation threads, making it useful for recalling and reusing prompts across different conversations or model variants. The prompt history list is accessible from the input area and supports a simple search filter for longer prompt libraries.
Conversation export
Signed-in accounts can export individual conversations from the Zhipu AI chat interface in plain text or markdown format. The export option appears in the conversation action menu alongside rename and delete controls. Markdown export preserves code blocks, headers, and list formatting from model responses; plain text export strips formatting for compatibility with downstream tools that do not parse markdown.
System prompt field
The system prompt field in Zhipu AI chat accepts a persistent instruction that frames the entire conversation. It is exposed in the conversation settings panel rather than inline in the message thread. Changes to the system prompt take effect from the next turn; they do not retroactively alter prior responses in the thread.
Browser support for Zhipu AI chat
The interface is tested on current versions of major browsers — streaming responses depend on a Fetch streaming capability that all modern browsers support.
The Zhipu AI chat interface runs on current versions of Chrome, Firefox, Safari, and Edge. Streaming responses — where model output appears token-by-token as it is generated — require a browser that supports the Fetch streaming API. All modern desktop and mobile browsers released after 2022 satisfy this requirement. Internet Explorer is not supported; legacy Edge (EdgeHTML engine) is also unsupported. Mobile browsers on iOS Safari and Android Chrome both work; the interface adapts to narrower viewports with the sidebar collapsing to a hamburger menu.
JavaScript must be enabled; the interface is a single-page application that does not render meaningfully without it. For users who require a no-JavaScript fallback, the Zhipuai API is the appropriate alternative for programmatic access without a browser-rendered UI.
Accessibility profile
The Zhipu AI chat interface meets the core practical accessibility requirements for keyboard and screen reader users, with some caveats on dynamic content announcements.
Keyboard navigation works throughout the main application: Tab moves focus through interactive elements in a logical order, Enter activates buttons and links, and the text input field receives focus automatically when a conversation opens. The model picker and conversation history list are both keyboard-navigable. Focus is managed correctly after model switching — the interface moves focus to the first element of the new conversation rather than leaving it stranded on the picker control.
Colour contrast ratios in the default theme meet WCAG 2.1 AA requirements for normal-sized text across the main interface elements. The streaming response area is announced to screen readers as content arrives, though the per-token update frequency can produce noisy live region behaviour on some screen reader and browser combinations. For accessibility-sensitive deployments, waiting for the full response to complete before reading is a more reliable pattern than consuming the live stream. Research from W3C's Web Accessibility Initiative provides the full technical standard that informed these interface decisions.
Zhipu AI chat feature availability
Feature availability varies between guest sessions and signed-in accounts — a summary of the five key features across both access modes.
| Feature | Availability | Notes |
|---|---|---|
| Conversation history | Signed-in only | Guest sessions are not saved; closing the tab ends the conversation permanently. |
| Model picker | Signed-in (full); guest (limited) | Guests see the default model only; signed-in accounts see all current GLM variants. |
| Prompt history recall | Signed-in only | Searchable list of recent prompts across all conversations for the account. |
| Conversation export | Signed-in only | Plain text and markdown formats; available from the conversation action menu. |
| System prompt field | Signed-in only | Accessible via conversation settings panel; persists for the session duration. |
Zhipu AI chat — frequently asked questions
Five questions across three tabs covering the product, features, and accessibility of the Zhipu AI chat interface.
What is Zhipu AI chat?
Zhipu AI chat is the official web chat interface for the Z.ai platform, serving as the browser-based product surface for the GLM model family. It provides conversation history, model switching, prompt recall, and export features for signed-in accounts, all running on the same inference infrastructure as the Zhipuai API.
How is Zhipu AI chat different from the chat Z AI workflow page?
The chat Z AI workflow reference covers how to use the interface — system prompts, model switching habits, when to escalate to the API. This page focuses on the Zhipu AI chat product itself — its features, browser compatibility, and accessibility profile. Both reference the same underlying interface from different perspectives.
Does Zhipu AI chat support conversation export?
Yes. Signed-in accounts can export individual conversations from the Zhipu AI chat interface in plain text or markdown format. The export option appears in the conversation action menu alongside rename and delete controls. Markdown export preserves code blocks and list formatting; plain text strips it for downstream compatibility.
Which browsers work with the Zhipu AI chat interface?
The Zhipu AI chat interface runs on current versions of Chrome, Firefox, Safari, and Edge. Streaming responses require the Fetch streaming API, which all modern desktop and mobile browsers support. Internet Explorer and legacy EdgeHTML are not supported. Mobile browsers on iOS Safari and Android Chrome both work with an adaptive layout.
Is the Zhipu AI chat interface accessible?
The interface includes keyboard navigation throughout, focus management after model switching, and WCAG 2.1 AA colour contrast ratios in the default theme. Screen reader support is functional for the conversation history list and the main input field. Dynamic streaming content may produce high-frequency announcements on some screen reader combinations; waiting for the complete response before reading is the more reliable pattern.
Related Z.ai access surfaces and references
The Zhipu AI chat interface connects to several other surfaces in the Z.ai ecosystem — understanding the full picture reduces duplication of effort.
The Zhipu AI chat interface is one of three primary ways to interact with the GLM model family. The second is the chat Z AI workflow, which frames the conversational interface from a practical usage perspective rather than a product feature inventory. The third is the Zhipuai API, which gives programmatic access to the same models with full parameter control. Behind all three sits the Zhipu AI open platform — the BigModel console that manages accounts, billing, and API keys. For users who prefer working in a dedicated interface rather than a general-purpose chat surface, the Z AI chatbot integration patterns offer a way to embed the model into existing tools. Account access for all surfaces starts at the Zhipuai login page, where international email registration is supported without a Chinese phone number in most regions.