---
format: "aidr-story-markdown/v1"
id: "849e9d7ee3636336233888117b07454befcadcd236b9342accdf42665b65c840"
canonical_url: "https://aidr.today/849e9d7e?lang=en"
title: "Zai’s GLM-5.3 Boosts Coding Performance 50% to Top Open Weight Benchmarks"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-08-31T21:00:10.000Z"
category: "Models"
topics: ["open-source","coding","agent","benchmark"]
source_urls: ["https://huggingnews.com/ai/zais-glm-53-boosts-coding-performance-50percent-to-top-open-weight-bench-ff14feab","https://x.com/arena/status/2094440382440611935","https://x.com/dabit3/status/2094432496142467191","https://x.com/FactoryAI/status/2094465930982265012","https://x.com/FactoryAI/status/2094465943523225681","https://x.com/flock_io/status/2094312007357083923","https://x.com/DeepLearningAI/status/2094421042785665237"]
summary: "Zai deployed GLM-5.3 and GLM-5.3-Flash models to agent platforms including Devin CLI, Droid, and the FLock API. The open-weight GLM-5.3 model achieved a 50% gain in coding performance over GLM-5.2 on the Code Bench, with specific improvements in complex programming and long-horizon agent tasks. The GLM-5.3-Flash version focuses on efficiency using a 30T-token multimodal training corpus and a combination of sparse and linear attention. In the Agent Arena, the Flash model ranks 4th among open source models and 19th overall based on 9,000 real-world sessions, recording a $0.12 median cost per task."
---

# Zai’s GLM\-5\.3 Boosts Coding Performance 50% to Top Open Weight Benchmarks

> [Open the canonical story](<https://aidr.today/849e9d7e?lang=en>)

**Published:** 2026-08-31T21:00:10.000Z
**Category:** Models
**Topics:** open\-source, coding, agent, benchmark

## Summary

Zai deployed GLM\-5\.3 and GLM\-5\.3\-Flash models to agent platforms including Devin CLI, Droid, and the FLock API\. The open\-weight GLM\-5\.3 model achieved a 50% gain in coding performance over GLM\-5\.2 on the Code Bench, with specific improvements in complex programming and long\-horizon agent tasks\. The GLM\-5\.3\-Flash version focuses on efficiency using a 30T\-token multimodal training corpus and a combination of sparse and linear attention\. In the Agent Arena, the Flash model ranks 4th among open source models and 19th overall based on 9,000 real\-world sessions, recording a $0\.12 median cost per task\.

## Sources

- [Story source](<https://huggingnews.com/ai/zais-glm-53-boosts-coding-performance-50percent-to-top-open-weight-bench-ff14feab>)
- [Story source](<https://x.com/arena/status/2094440382440611935>)
- [Supporting source](<https://x.com/dabit3/status/2094432496142467191>)
- [Supporting source](<https://x.com/FactoryAI/status/2094465930982265012>)
- [Supporting source](<https://x.com/FactoryAI/status/2094465943523225681>)
- [Supporting source](<https://x.com/flock_io/status/2094312007357083923>)
- [Supporting source](<https://x.com/DeepLearningAI/status/2094421042785665237>)