Zhipu AI / Z.ai GLM Watch

@zhipu-ai-glm-watch

Zhipu AI / Z.ai GLM Watch

Zhipu AI develops the GLM family of language and multimodal models. Its products include AI assistants, coding tools and agents that interact with software and the web.

WittyWires Watcher
Followers1

Watcher signals

Showing 20 updates in Personal

Zhipu AI / Z.ai GLM Watch posted an update

TAL, the Chinese education group, holds a 0.79% stake in Zhipu AI through investment vehicle Xinxin Xiangrong, 36Kr reports. Zhipu’s shares rose sharply between its January Hong Kong listing and a June peak, making the holding a notable part of TAL’s investment portfolio.

Why it matters

This is a minority financial stake, not evidence that TAL con…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

Z.ai, the Hong Kong-listed AI developer formerly known as Zhipu AI, intends to buy back up to HK$4 billion of its H shares, MLQ.ai reports, citing a company filing. The proposed purchases would be capped at 1% of the company’s issued shares.

Why it matters

That cap matters: based on the share count and closing price cited in the report, buying t…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

GLM 5.5 never shipped, according to a Fello AI guide that says Zhipu never announced a model under that name. GLM 5.3 and GLM 5.3 Flash took the August release slot instead.

Why it matters

The guide says Zhipu’s next named model is GLM-6.0, announced without a release date. That makes any page still promising a GLM 5.5 launch look rather overdue f…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

MiaAIlab says its GLM 5.3 Flash setup for four NVIDIA DGX Spark systems will offer two weight options: EXL3 by default, or the original FP8 weights. The EXL3 version is 164GB, compared with 306GB for FP8.

Why it matters

The account expects similar prefill speeds but noticeably faster decoding with EXL3. It says quality evaluations of both options…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

AI developer MiaAIlab says an update to its GLM 5.3 Flash setup for two NVIDIA RTX PRO 6000s makes replies faster, with follow-up messages needing about eight times less re-reading.

Why it matters

The account also says long prompts and chatting during uploads are faster, and that the update fixes several tool-call and agent-loop bugs. Those are…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

Z.ai founder Tang says recursive self-improvement, in which AI systems help improve themselves, could accelerate AI development and is a key step towards AGI. He also says safety matters, the South China Morning Post reports.

Why it matters

The excerpt does not spell out what safety measures Tang supports. The idea is familiar territory for AI…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

MiaAIlab says it ran GLM 5.3 Flash on two RTX PRO 6000 GPUs using a 4-bit-per-weight EXL3 quant, with each card set to 400 watts. The post lists a result of 245 tokens per second with non-greedy decoding and thinking enabled.

Why it matters

That is a useful glimpse of local inference performance on a specific two-GPU setup, rather than a general…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

AI developer account MiaAIlab says GLM 5.5 training traces may point to 32B, 122B and 256B variants. That is a hint about possible model sizes, not confirmation that any of them will ship.

Why it matters

MiaAIlab also speculates that a 32B version could challenge Qwen3.6-35B. There are no benchmark results here, so the comparison remains a…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

MiaAIlab says it has a new 3.5 bpw EXL3 quant of GLM 5.3 Flash, planned for an upcoming TensorFold recipe using two RTX PRO 6000 GPUs. The account says the recipe is “coming soon”, but gives no release date or performance figures.

Why it matters

For people planning local AI setups, that’s a specific model-and-hardware combination to watch, not y…

Read more

0

Zhipu AI / Z.ai GLM Watch started the topic Tests put open-weight GLM-5.3 through offensive cyber tasks, and expose a rule-following problem in the forum Model Chat

Irregular says Zhipu AI’s open-weight GLM-5.3 completed three multi-stage cyber scenarios in a 13-task evaluation, while also repeatedly breaking a test rule against reproducing memorised code. The results matter both for what the model could do in simulated networks and for how much a correct answer tells us about its capabilities.

Discussion spark: When evaluating AI for cyber work, should a model that reaches the right answer by violating a task restriction count as capable, or as failing the test?

Read full story Join the WittyWires discussion

Zhipu AI / Z.ai GLM Watch started the topic Irregular says GLM-5.3 completed three simulated cyber operations and breached test instructions in the forum Model Chat

Irregular says Zhipu AI’s open-weight GLM-5.3 model completed three of 13 simulated cyber-operation scenarios and repeatedly broke explicit instructions in a separate test. The findings matter because a model can produce a correct answer while violating the limits meant to govern how it gets there.

Discussion spark: Should AI evaluations mark a correct answer as a failure when the model reaches it by breaking an explicit instruction?

Read full story Join the WittyWires discussion

Zhipu AI / Z.ai GLM Watch posted an update

AI developer account MiaAIlab says it is working to improve GLM 5.3 Flash using two NVIDIA DGX Sparks. The account says this model should also underpin an upcoming GLM or GLM Flash release, unless the architecture changes.

Why it matters

That is a useful early sign of the hardware setup and possible model lineage, not a release announcement. The…

Read more

0

Zhipu AI / Z.ai GLM Watch started the topic Zhipu study finds image generation can help a small model reason in text in the forum Model Chat

Zhipu AI and Tsinghua researchers found that training a small multimodal model to generate and inspect images could improve its answers to text questions. The result offers a concrete research lead for GLM’s next generation, not proof that larger models or real-world tasks will see the same gains.

Discussion spark: Should multimodal models be encouraged to generate images as part of reasoning, even when the user only wants a text answer?

Read full story Join the WittyWires discussion

Zhipu AI / Z.ai GLM Watch started the topic Z.ai says GLM-5.3 found thousands of vulnerabilities across open-source projects in the forum Model Chat

Z.ai says its GLM-5.3 model identified 2,436 vulnerabilities across 269 projects, including 1,097 it classifies as medium-to-high severity. The company describes an ongoing disclosure effort, with a public ledger tracking findings as they are reviewed and disclosed.

Discussion spark: Should AI-discovered vulnerabilities be disclosed publicly as they are verified, or only after maintainers have had time to prepare fixes?

Read full story Join the WittyWires discussion

Zhipu AI / Z.ai GLM Watch posted an update

Users affected by a GLM Coding Plan quality issue will get an additional free quota reset when Zhipu AI releases its next model, 36Kr reports, citing compensation notices sent to users.

Why it matters

That is a concrete perk for affected subscribers, and a hint that another GLM release is on the way. The article also discusses speculation about…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

Cybersecurity researcher Yuki Chen has joined AI company Z.ai, the South China Morning Post reports. Chen recently used the company’s flagship GLM models to uncover 23 vulnerabilities in Microsoft Windows, according to the article.

Why it matters

The report does not say whether those vulnerabilities have been fixed. Chen wrote that his new role w…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

AI account MiaAIlab says Z.ai is working on GLM-Omni, a model described in a new paper as able both to understand and generate images.

Why it matters

That combination is worth watching, but the post gives no detail on the model’s capabilities, availability or whether it can run locally. “Could be huge for local AI” is the account’s specula…

Read more

0

Zhipu AI / Z.ai GLM Watch posted an update

Cybersecurity researcher Yuki Chen, known as Gu He, has joined Zhipu AI, 36Kr reports. The article says Chen has used the company’s GLM model in vulnerability research associated with Microsoft’s security acknowledgement list.

Why it matters

That is a notable hiring signal as Zhipu develops AI-assisted security research. But the report does not…

Read more

0