Z.AI unveils Ox Alpha GLM model and plans to release its weights

GLM-5.3-Flash, also known as ox-alpha, is out now! Zai's GLM-5.3-Flash is a new 320B parameter multimodal open model with 18B active parameters. GLM-5.3-Flash approaches Claude 4.8 Opus on coding and agentic benchmarks. GGUF coming soon!

A mystery model showed up on OpenRouter for free and developers spent days debating whether it was Google, Microsoft or Cursor because its coding performance was so good. Turns out it was Z.ai’s next GLM model from China and the weights are being released. Model commoditization is happening fast.

Arguably the most interesting thing about GLM 5.3 Flash (Ox Alpha) is that its inference has been entirely on Chinese-made AI chips. I always thought that American export restrictions would backfire and just encourage Chinese companies to develop their own alternatives. That indeed has happened now.

z.AI acquired a 60% stake in Zhongke Jiahe, a firm that likely played an instrumental role in running Ox-Alpha at production scale The firm specializes in heterogeneous compute, so it’s likely that Huawei is just one of many “Chinese chips” used to train & inference www.stcn.com/article/deta...

While you all argue about which US AI exec is saying which dumb thing or the etymology of cyberslurs, China's Long AI March continues. I've been using GLM 5.3 Flash (as ox-alpha) all week. It's real and it's my friend. China is 100% pacing the US frontier labs.

Looks like @pekka.bsky.social got this one right! Ox Alpha now confirmed to be glm-5.3.flash.

Holy crap. GLM-5.3-Flash also releases - 321B You'll need approximately double the hardware of the Qwen3.8 Flash model to run. Users of "Ox Alpha", its pre-release code name, positively compared it to Anthropic's Fable.

Zhipu AI's new Ox Alpha GLM model, strong in coding & agents, now open-source! 🚀 Supports text, image, video. Topped OpenRouter charts! Free for a week. #AI #LLM #OpenSource

On Zhipu's incredible new model, GLM 5.3 Flash, which was the previously unidentified frontier model Ox-Alpha, and what this means for the commodificaiton of intelligence. www.strangeloopcanon.com/p/who-wins-a...

Z.ai has confirmed that Ox Alpha is a new GLM-series model and says the weights will be released. Until they are, everything published about its behavior comes from an endpoint, not from anything anyone outside Z.ai can run and inspect. #dev

than GLM 5.2 because GLM 5.2 has enough of an instinct about the stupid way CIDCO (local urban planning corporation) handles digitisation that it could be more token efficient, meanwhile ox alpha was just as bad as deepseek v4 flash; it's why I said ox alpha "lacks world knowledge" in a past post

...Ox Alpha was actually GLM Flash unironically? And it's super cheap? z.ai/blog/glm-5.3...

So ox alpha is glm 5.3 flash, and they’re saying it was all served with Chinese chips Extremely impressive z.ai/blog/glm-5.3...

China, getting the job done. ' This has led to increased concerns about China potentially overtaking Western powers like the US in the AI race by consistently rolling out highly sophisticated and often unclaimed models. '

New Mysterious AI Model dubbed Ox Alpha with Free 100 Trillion Tokens a Day for Coders: cybersecuritynews.com/new-mysterio...