<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/QCRtwG9jI" -->

---
title: Z.ai revealed that the anonymously tested ox-alpha was...
description: Z.ai released GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series, with 320B total parameters and 18B active parameters. It uses a hybrid...
canonical: https://daily.dev/posts/z-ai-revealed-that-the-anonymously-tested-ox-alpha-was-glm-5-3-flash-a-320b-parameter-moe-model-wit-qcrtwg9ji
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Z.ai revealed that the anonymously tested ox-alpha was GLM-5.3-Flash, a 320B-parameter MoE model with 18B active parameters that approached Claude Opus 4.8 on coding and agentic benchmarks. | daily.dev
og:description: Z.ai released GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series, with 320B total parameters and 18B active parameters. It uses a hybrid...
og:url: https://daily.dev/posts/z-ai-revealed-that-the-anonymously-tested-ox-alpha-was-glm-5-3-flash-a-320b-parameter-moe-model-wit-qcrtwg9ji
og:image: https://api.daily.dev/og/posts/QCRtwG9jI.png
og:image:alt: Z.ai revealed that the anonymously tested ox-alpha was GLM-5.3-Flash, a 320B-parameter MoE model with 18B active parameters that approached Claude Opus 4.8 on coding and agentic benchmarks.
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Z.ai revealed that the anonymously tested ox-alpha was GLM-5.3-Flash, a 320B-parameter MoE model with 18B active parameters that approached Claude Opus 4.8 on coding and agentic benchmarks.

**[Andrew M](https://daily.dev/sources/dup6ye036hupdshygexyu)** · [@andrewma](https://daily.dev/andrewma) · 9 upvotes · 2 comments

## Summary

Z.ai released GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series, with 320B total parameters and 18B active parameters. It uses a hybrid sparse and linear attention architecture plus Manifold-Constrained Hyper-Connections to cut long-context serving costs, and was trained on a 30T-token multimodal corpus. It outperforms GLM-5.2 across coding and agentic benchmarks at one-tenth the price while approaching Claude Opus 4.8, and was served entirely on Chinese AI chips achieving a 3x serving performance improvement. Weights are available on HuggingFace with support for SGLang, vLLM, and TokenSpeed, and it has been rolled out to all GLM Coding Plan users with 3x the quota of GLM-5.3.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://z.ai/blog/glm-5.3-flash>

## Community discussion

Top comments from developers on daily.dev.

**@jepudev2** · 2 upvotes

> it was really good, i had it running alongside gpt terra and it fixed problems terra couldnt solve

**@agustinbarrientos** · 0 upvotes

> I bet the same-hardware throughput result matters more than another benchmark table once the serving recipe is independently reproduced

## Similar posts on daily.dev

- [GLM 5.3 FlashX now available on AI Gateway](https://daily.dev/posts/glm-5-3-flashx-now-available-on-ai-gateway-9mhh2yigq) · Vercel · 0 upvotes · 0 comments
- [GLM-5.2: Built for Long-Horizon Tasks](https://daily.dev/posts/glm-5-2-built-for-long-horizon-tasks-stmnmwjok) · Hugging Face · 33 upvotes · 4 comments

---

Tags: [#llm](https://daily.dev/tags/llm), [#mixture-of-experts](https://daily.dev/tags/mixture-of-experts)

[View this post on daily.dev](https://daily.dev/posts/z-ai-revealed-that-the-anonymously-tested-ox-alpha-was-glm-5-3-flash-a-320b-parameter-moe-model-wit-qcrtwg9ji)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/z-ai-revealed-that-the-anonymously-tested-ox-alpha-was-glm-5-3-flash-a-320b-parameter-moe-model-wit-qcrtwg9ji","headline":"Z.ai revealed that the anonymously tested ox-alpha was GLM-5.3-Flash, a 320B-parameter MoE model with 18B active parameters that approached Claude Opus 4.8 on coding and agentic benchmarks.","text":"Shared: GLM-5.3-Flash: Frontier Intelligence, Flash Cost","url":"https://daily.dev/posts/z-ai-revealed-that-the-anonymously-tested-ox-alpha-was-glm-5-3-flash-a-320b-parameter-moe-model-wit-qcrtwg9ji","datePublished":"2026-08-27T16:16:03.848Z","dateModified":"2026-08-27T16:55:19.373Z","author":{"@type":"Person","name":"Andrew M","url":"https://daily.dev/andrewma","image":"https://avatars.githubusercontent.com/u/102819214?v=4","description":"full time overthinker\n","interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"EndorseAction"},"userInteractionCount":3930}},"image":"https://media.daily.dev/image/upload/s--foaA6JGU--/f_auto/v1722860399/public/Placeholder%2004","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":9},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":2}],"sharedContent":{"@type":"WebPage","url":"https://api.daily.dev/r/fMxTjGcaJ"},"comment":[{"@type":"Comment","text":"it was really good, i had it running alongside gpt terra and it fixed problems terra couldnt solve","datePublished":"2026-08-27T22:36:40.195Z","url":"https://daily.dev/posts/QCRtwG9jI#c-df5W189hc","author":{"@type":"Person","name":"jepudev","url":"https://daily.dev/jepudev2","image":"https://avatars.githubusercontent.com/u/71714253?v=4"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2}},{"@type":"Comment","text":"I bet the same-hardware throughput result matters more than another benchmark table once the serving recipe is independently reproduced","datePublished":"2026-08-31T23:39:08.895Z","url":"https://daily.dev/posts/QCRtwG9jI#c-NiG0IL1Ui","author":{"@type":"Person","name":"Agustin Barrientos","url":"https://daily.dev/agustinbarrientos","image":"https://media.daily.dev/image/upload/s--5ayxQnqn--/f_auto/v1788281802/avatars/avatar_wQYYVe5Tbj0NJ7C7qPoa8?_a=BAMAMicg0"}}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/dup6ye036hupdshygexyu","name":"Andrew M"}}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Andrew M","item":"https://daily.dev/sources/dup6ye036hupdshygexyu"},{"@type":"ListItem","position":3,"name":"Z.ai revealed that the anonymously tested ox-alpha was GLM-5.3-Flash, a 320B-parameter MoE model with 18B active parameters that approached Claude Opus 4.8 on coding and agentic benchmarks."}]}
```

