Introducing the New GLM-5.3 Flash Model
Zhipu has officially launched the GLM-5.3 Flash model, showcasing impressive performance results.
Introducing the New GLM-5.3 Flash Model
Zhipu has officially introduced and launched the GLM-5.3 Flash. This model has confirmed its identity as the anonymous Ox Alpha, which previously appeared in OpenCode and OpenRouter.
The official release is more powerful and noticeably more stable.
According to Z.ai, GLM-5.3 Flash scores 84.3 in Terminal-Bench 2.1, nearly catching up to Claude Opus 4.8 with a score of 85.0.
It outperforms the significantly larger GLM-5.2 in all six published tests and costs about 10 times less to use.
- Input — $0.15 per million tokens.
- Output — $0.50.
- Cached input — $0.03.
Open weights, MIT license, built-in multimodality, and context for 1 million tokens.
Why it matters
AnalysisThis model represents a significant advancement in artificial intelligence technology, demonstrating high performance at an accessible price. It could serve as a foundation for new applications across various fields.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
MIT researchers show GPT-6 Astra can improve other AI models by editing their weights
In an experiment, Astra raised Qwen3.5-0.8B's accuracy on MATH-500 from 7.9% to 48.1% by changing only 0.016% of the model's weights.
JSON — Structured Data for Programs, APIs, and AI Agents
A complete hands-on guide to JSON: syntax specifications, serialization via parse and stringify, REST API consumption, JSON Schema validation, and structured outputs for LLMs.
DOCX and XLSX Formats: When AI Needs to Deliver Production-Ready Documents
A comprehensive hands-on guide to generating and processing DOCX documents and XLSX workbooks with AI: comparing against Markdown and CSV, prompt engineering, formula design, and code automation.
Binance publishes investigation into $6.3 billion theft
Binance’s investigation links the theft to a modified Homebrew download, a fake interview and abuse of an internal AI tool.