GLM-5.5 Release Date: Not Announced Yet, What to Use Now
Z.ai has never named GLM-5.5. Here is where the talk started, what can be checked, and which GLM model to call today.
On this page
- Can you use GLM-5.5 today?
- What has Z.ai confirmed?
- What Z.ai has not said
- Where do the GLM-5.5 rumors come from?
- If you need a GLM model this week
- Which GLM models can you call today?
- Frequently asked questions
- Is GLM-5.5 out yet?
- When will GLM-5.5 be released?
- How many parameters will GLM-5.5 have?
- What is the latest GLM model?
- Is GLM-5.3-Prime a new model?
- Sources checked
Z.ai (Zhipu AI) has not announced GLM-5.5. As of November 1, 2026, its docs, price page, release log, Hugging Face and GitHub accounts name no GLM-5.4, GLM-5.5 or GLM-6. The newest flagship is GLM-5.3, released August 18, 2026.
Short answer
- Not announced. No official page or account checked on November 1, 2026 lists GLM-5.5.
- The trillion-parameter and "2.5–3.5T" figures come from Baidu Baike, CometAPI and an X post. Z.ai has confirmed none of them.
- Use GLM-5.3 for coding and agent work, GLM-5.3-Flash for image, video or file input, and GLM-5.3-FlashX when about 200 tokens per second matters.
Can you use GLM-5.5 today?
No. Z.ai has published nothing to call under that name, and none of the pages below, checked on November 1, 2026, goes past the GLM-5.3 family.
- Z.ai's release notes end on August 26, 2026, with GLM-5.3-Flash.
- Z.ai's price page lists three models in its "Latest Models" table: GLM-5.3, GLM-5.3-Flash and GLM-5.3-FlashX.
- Z.ai's models overview names GLM-5.3 as the flagship and GLM-5.3-Flash as the multimodal model, both with a 1M-token context window.
- The zai-org account on Hugging Face carries no GLM-5.4, GLM-5.5 or GLM-6 repository. Its newest GLM uploads are the GLM-5.3 family, created August 25, 2026.
- Of the 30 most recently updated repositories in the zai-org account on GitHub, none is a GLM-5.3, GLM-5.4 or GLM-5.5 repository; the GLM model repository there is GLM-5.
- OpenRouter's model list has no z-ai/glm-5.4 or z-ai/glm-5.5. Its newest Z.ai listings are z-ai/glm-5.3-prime, added September 23, 2026, and z-ai/glm-5.3-flashx, added September 18, 2026.
- Z.ai's GLM-5.3 blog post is dated August 14, 2026, and the addresses /blog/glm-5.4 and /blog/glm-6 return "not found".
- Zhipu's China platform docs list the same three models and date GLM-5.3 August 19, 2026, one day after the international log.
A search for "GLM-5.4" and "GLM-5.5" in every spelling, from "GLM 5.5" to "glm5.5", across all of those pages returned 0 hits. "GLM-6" matches only old names such as ChatGLM-6B.
What has Z.ai confirmed?
Everything Z.ai has confirmed belongs to the GLM-5.3 family. The table pulls from Z.ai's GLM-5.3 docs and GLM-5.3-Flash docs, both checked November 1, 2026.
| Item | What Z.ai says | Date |
|---|---|---|
| Newest flagship | "GLM-5.3 is Z.ai's latest flagship model" | August 18, 2026 |
| Newest dated release | GLM-5.3-Flash, first natively multimodal GLM-5 | August 26, 2026 |
| Fast version | GLM-5.3-FlashX "now live", 200 tokens/s | Undated |
| What comes next | A research direction only | Undated |
| GLM-5.5 mentions | 0 hits | November 1, 2026 |
GLM-5.3 takes text only, with a 1M-token context window and up to 128K output tokens; the model code is glm-5.3. Reasoning is always on: reasoning_effort accepts low, high or max, and an app that used to send thinking.type: "disabled" should send reasoning_effort set to low. GLM-5.3-Flash adds video, image and file input, with 320B total parameters and 18B active, and keeps the same 1M context and 128K output. Its faster sibling runs under glm-5.3-flashx. Z.ai's GLM-5.3 blog post is dated August 14, 2026; the release notes date the model August 18.
The one forward-looking line on the GLM-5.3 page is a research note rather than a model: "making environment generation and verification more autonomous is one of the next steps".
What Z.ai has not said
Seven pieces of a GLM-5.5 plan are still missing:
- Whether a GLM-5.5 is coming at all.
- A date, or a launch window.
- Its size, including the parameter count.
- A price.
- A context window and an output limit.
- Whether the weights will be open, and under which licence.
- Whether it will run on the GLM Coding Plan.
Where do the GLM-5.5 rumors come from?
Z.ai has not confirmed any of them. Each lead is named below with its date.
- Baidu Baike, English entry "GLM-5.5" (no page date): "GLM-5.5 is a trillion-parameter Large Language Model that J.P. Morgan … predicts Zhipu AI will release in August". Its two references, a mydrivers.com article and a Baijiahao post, are both dated June 22, 2026; they carry no links, and neither cites a J.P. Morgan document. August passed with GLM-5.3 instead.
- CometAPI, "GLM-5.5 expected specs", published August 19 and updated September 3, 2026: "Reuters has identified GLM-5.5 as a later roadmap milestone, but Z.ai has not published an exact launch commitment." The Reuters article it cites, dated June 25, 2026, refused our request with HTTP 401, so we could not check that line.
- CometAPI, "GLM-5.5 leak", published October 25 and updated October 26, 2026: "As of late October 2026, Z.ai (Zhipu AI) has not officially released or confirmed GLM-5.5." It passes on leaks of a next flagship "in the 2.5–3.5 trillion range" and adds that nothing public supports a specific trillion-parameter claim for GLM-5.5.
- X, @NFT_Chen, October 21, 2026: the post says leaked "expert minutes" on GLM-5.4 and GLM-5.5 put Zhipu's next base model at 2.5–3.5T parameters.
On timing, CometAPI's October 25 post says speculation has moved from August to "late October/early November 2026", citing gaps of about 54–70 days between past releases and a compensation notice that teased a "new model". For that tease it cites the ZCode changelog, which on November 1, 2026 listed only GLM-5.3-era product updates. Z.ai's own blog dates show the rhythm: GLM-5 on February 12, GLM-5.1 on April 7, GLM-5.2 on June 16 and GLM-5.3 on August 14, 2026, a new GLM-5.x version about every two months. Those gaps were 54, 70 and 59 days, and by November 1, 2026, 79 days had passed since the GLM-5.3 post, longer than any earlier gap. Past gaps promise nothing about the next date.
In open-source inference engines, a GitHub search for "glm-5.5" finds 0 issues or pull requests in Hugging Face transformers and SGLang; the 5 hits in vLLM are unrelated text. OpenRouter also lists a z-ai/glm-5.3-prime, added September 23, 2026 and described as "the high-speed variant of Z.ai's GLM-5.3", and Z.ai's own pages leave that name out.
Developers are guessing too. "MiniMax M3 Pro is also coming, then DeepSeek-v4-Pro GA, then GLM5.5", ycui7 wrote on Hacker News on August 3, 2026.
If you need a GLM model this week
Pick by job.
| If you need | Use now |
|---|---|
| Z.ai's flagship for coding and agents | GLM-5.3 |
| Image, video or file input, or the lowest price | GLM-5.3-Flash |
| Speed, about 200 tokens per second | GLM-5.3-FlashX |
| The GLM Coding Plan | GLM-5.3 or GLM-5.3-Flash; the plan leaves out FlashX for now |
| To move to GLM-5.5 on day one | Keep the model ID in config, so the swap is one string |
Z.ai's GLM Coding Plan page (checked November 1, 2026) says the plan "now supports the latest GLM-5.3 and GLM-5.3-Flash models for all users (Max, Pro, and Lite)". The 200 tokens per second figure is Z.ai's own; our GLM-5.3-FlashX review weighs that speed against the price gap with GLM-5.3-Flash.
Waiting on Alibaba's next model too? Our Qwen 4 release date page runs the same checks for Qwen 4.
Which GLM models can you call today?
Three, all from the GLM-5.3 family.
All three are also on OmniaKey's OpenAI-compatible API: GLM-5.3, GLM-5.3-Flash and GLM-5.3-FlashX. Each model page shows the current rate, and pricing lists every model. For how GLM-5.3 does on coding, see GLM-5.3 vs GLM-5.2 for coding.
Frequently asked questions
Is GLM-5.5 out yet?
No. As of November 1, 2026, Z.ai's release notes, price page and model docs stop at the GLM-5.3 family, and so do its Hugging Face account and OpenRouter's model list.
When will GLM-5.5 be released?
Z.ai has given no date. The "late October/early November 2026" window in CometAPI's October 25, 2026 post rests on the gaps between earlier releases; by November 1, 2026, 79 days had passed since the GLM-5.3 blog post, longer than any earlier gap.
How many parameters will GLM-5.5 have?
Z.ai has published no size. The trillion-parameter figure comes from a Baidu Baike entry that cites no J.P. Morgan document, and the "2.5–3.5T" range from an X post by @NFT_Chen on October 21, 2026. CometAPI itself says nothing public supports a specific trillion-parameter figure.
What is the latest GLM model?
GLM-5.3 is Z.ai's latest flagship, released August 18, 2026. GLM-5.3-Flash followed on August 26, 2026, the newest entry in Z.ai's release notes. GLM-5.3-FlashX is live too, at 200 tokens per second; Z.ai leaves it undated, and OpenRouter added it on September 18, 2026.
Is GLM-5.3-Prime a new model?
Z.ai has not announced one. OpenRouter added z-ai/glm-5.3-prime on September 23, 2026 and describes it as "the high-speed variant of Z.ai's GLM-5.3", but no Z.ai page, price table or release log lists it. Treat it as an OpenRouter listing until Z.ai names it.
Sources checked
- Z.ai release notes
- Z.ai API pricing
- Z.ai models overview
- Z.ai GLM-5.3 documentation
- Z.ai GLM-5.3-Flash documentation
- Z.ai GLM Coding Plan
- Z.ai blog: GLM-5.3
- Z.ai blog: GLM-5.2
- Zhipu AI China platform release log
- Hugging Face: zai-org
- GitHub: zai-org
- OpenRouter models
- Hacker News comment by ycui7, August 3, 2026
Updated November 1, 2026, 00:18 UTC. We will revise the page when Z.ai's release notes or price page list GLM-5.5.
Models in this article
Related articles
GLM-5.3-FlashX Review
FlashX sells lower waiting time, not a proven new intelligence tier, so its roughly 2.5x premium belongs only on latency-sensitive paths.
GLM-5.3 vs GLM-5.2 for Coding
GLM-5.3 keeps the GLM-5.2 base model and list price, but changes the coding decision through stronger post-training, new effort controls, and a breaking thinking-mode migration.
GLM-5.2 API Pricing
Z.ai lists GLM-5.2 at $1.40 input, $0.26 cached input, and $4.40 output per million tokens.