Z.ai · Model guide
GLM-5.3-Flash
The first natively multimodal model in the GLM-5 series. A 320B-parameter mixture of experts that activates 18B per token, it combines sparse and linear attention and folds vision directly into the coding loop, so it can inspect rendered output, interfaces and documents while it works. It targets frontend and long-horizon agentic engineering as well as Office and financial document workflows, across a 1M-token context.
What this model is good at
What the vendor positions it for, and which of our models to reach for instead.
VendorZ.AI positions it as the first natively multimodal model in the GLM-5 series, reaching stronger intelligence than GLM-5.2 through a hybrid sparse and linear attention architecture.source
Visual coding loop · Vision is integrated natively into the coding loop, so the model can inspect its own rendered output and interaction feedback and refine frontend, game and 3D work.(vendor claim)source
Long-horizon agentic engineering · The official comparison reports Terminal-Bench 2.1, DeepSWE v1.1 and Toolathlon Verified well above GLM-5.2.source
Office and financial documents · The vendor extends it to Office, financial research and professional document work, delivering PPTX, PDF, DOCX and XLSX artefacts.(vendor claim)source
- z-ai/glm-5.3sibling —Choose GLM-5.3 for the text-only flagship tuned for complex software engineering and cybersecurity work.
- z-ai/glm-5.2predecessor —Use GLM-5.2 when the application must be able to turn thinking off.
Pricing and billing
Billed per token. Cached input is charged at the cache-read rate.
| Price / 1M tokens | You pay |
|---|---|
| Input | $ 0.15$ 0.075 |
| Output | $ 0.50$ 0.25 |
| Cache read | $ 0.030$ 0.015 |
| Cache write | $ 0.15$ 0.075 |
+ 2.0K × $ 0.25
Capabilities and limits
What CrossModel guarantees across every route this model can take right now.
- Context window
- 1.0M tokens
- Max output
- 128.0K tokens
- Input / output modalities
- Text + Image + Video → Text
- Streaming
- Supported
- Tool calling
- Supported
- Structured output (JSON)
- Supported
- Image input
- Supported
- Thinking mode
- Supported· always on
- Reasoning effort
- low · high · max
- Available endpoints
- /v1/chat/completions · /v1/responses · /v1/messages
Published benchmarks
Scores the vendor reported, with the evaluation setup each one came from.
- Agents' Last Exam (CLI)score·Z.AI · 2026-08-26effort maxClaude Codesetup not fully disclosed
按官方 ALE 评测协议、由官方评测器打分；1M 上下文，Tool Search 关闭。
26.3% - AutomationBench v1.0.6score·Z.AI · 2026-08-26setup not fully disclosed
评测于 AutomationBench v1.0.6，并合入 PR #13 对 null 类型处理的修复；与未标版本的 AutomationBench 结果不可直接比较。
48.8%
- 1773
- AA Intelligence Index v4.1.1score·Z.AI · 2026-08-26enabledsetup not fully disclosed
Artificial Analysis 的九项复合指数，不是百分比；Z.AI 在发布页转载该分数，未披露各分项的可复现设置。
57 - Humanity's Last Exam (with tools)accuracy·Z.AI · 2026-08-26setup not fully disclosed
全集评测，最大上下文 300K 并启用上下文管理策略，判分模型为 GPT-5.6-luna (medium)；官方未列出完整工具集合。
55.3%
Vendor-reported numbers, not CrossModel measurements. Scores are only comparable when the benchmark version, metric and evaluation setup match, so nothing here is averaged or ranked.
Use it in your tools
Point the base URL at CrossModel and paste this model ID — every tool below has a setup guide.
Related models
Other tiers from the same vendor, plus similarly priced alternatives.
Frequently asked questions
What is GLM-5.3-Flash?
How much does GLM-5.3-Flash cost?
Does GLM-5.3-Flash support tool calling and structured output?
Which endpoint do I call?
Can I try it without writing code?
Chat first, integrate later
Chat first, then wire it in once you like the answers.