Is GLM-5.4 Out? What's Confirmed, and What You Can Call Today
As of 23 Nov 2026, Asia/Shanghai, GLM-5.4 is not officially released and is not on SandBase. Call GLM-5.3, GLM-5.3-Flash, or GLM-5.3-FlashX today.
On 23 November 2026, Asia/Shanghai, GLM-5.4 is not an official product. Z.ai and Zhipu have not announced it. The release notes still end at GLM-5.3 and GLM-5.3-Flash. Official model pages, the official blog path, and both pricing tables have no GLM-5.4 row, and there is no official API ID, price, parameter count, license, or weights. SandBase has no GLM-5.4 card. The catalog does list a different name, z-ai/glm-5.5, and that card says Coming soon. It is not callable either. The watch post for that name is Is GLM-5.5 Out? What’s Confirmed, and What You Can Call Today.
If you need a request that works today, the live IDs are z-ai/glm-5.3, z-ai/glm-5.3-flash, and z-ai/glm-5.3-flashx. This page is a status check for the search “GLM 5.4”. It is not a launch note, and it does not replace the existing GLM-5.3 and Flash writeups linked below.
SandBase catalog titles often print “GLM 5.3” and “GLM 5.5” without a hyphen. Official spelling in the running copy stays hyphenated: GLM-5.4, GLM-5.5, GLM-5.3, GLM-5.3-Flash, GLM-5.3-FlashX. There is no SandBase title for GLM-5.4, because there is no card.
Key takeaway
- Checked 23 Nov 2026, Asia/Shanghai. No official GLM-5.4 announcement, model card, API ID, price, context window, parameter count, license, or weight release.
- Official changelogs still top out at GLM-5.3-Flash on 26 Aug 2026 and GLM-5.3 on 18 Aug 2026 (Z.ai) or 2026-8-19 (BigModel). GLM-5.3 is the latest flagship. GLM-5.3-Flash is the native multimodal sibling. GLM-5.3-FlashX is the high-speed Flash tier.
- SandBase has no GLM-5.4 card.
z-ai/glm-5.5is a separate coming-soon listing, not an official 5.4, and not a working call. The Facts line and the 131.1K max-output figure on that card are catalog placeholders.- Callable today on SandBase: GLM-5.3 at 25% off ($1.05 / $3.30 / cache read $0.19 per 1M), GLM-5.3-Flash at 40% off ($0.09 / $0.30 / cache read $0.02), and GLM-5.3-FlashX at $0.37 / $1.25 / cache read $0.09, with no discount stated. Route:
POST https://api.sandbase.ai/v1/chat/completions.
Short answer: not officially announced, and not callable
The search is “GLM 5.4”. There is no official product page to match that name. Do not treat glm-5.4 as a shipping vendor SKU. SandBase does not have a catalog ID for it.
| Claim | Grade | What the 23 Nov 2026 check found |
|---|---|---|
| Official GLM-5.4 announcement, model page, API ID, or price | Confirmed absent | Z.ai and BigModel release notes have zero matches. Both pricing tables have no 5.4 row. The official model and blog paths returned HTTP 404. |
| Current official flagship | Confirmed | GLM-5.3, vendor ID glm-5.3. Z.ai docs call it the latest flagship. Text only. 1M context. Official max output 128K. |
| Native multimodal sibling and fast tier | Confirmed | GLM-5.3-Flash (glm-5.3-flash) and GLM-5.3-FlashX (glm-5.3-flashx), listed 26 Aug 2026. |
| A public weight repo named glm-5.4, or an OpenRouter glm-5.4 / glm-5.5 id | Confirmed absent | Hugging Face model search for glm-5.4 returned 0 repos. OpenRouter’s model list has no glm-5.4 id and no glm-5.5 id. |
| SandBase has a GLM-5.4 card you can call | Confirmed absent | Both catalog paths returned HTTP 404. There is a coming-soon card for a different string, z-ai/glm-5.5. That card is not a 5.4 launch. |
| A late-November release date, a 3T parameter count, or GLM-6.0 as the next official name | Reported / Unknown | Not on the official pages in this check. Not used below for names, IDs, or prices. |
The 200 pages behind that table are the Z.ai release notes, the BigModel release log, Z.ai pricing, and BigModel pricing. https://docs.z.ai/llms.txt lists the GLM-5.3, GLM-5.3-Flash/FlashX, and GLM-5.2 docs and does not list a GLM-5.4 or GLM-5.5 page. The chat app title on chat.z.ai still reads “Z.ai - Advanced AI Chatbot & Agent powered by GLM-5.3-Flash”. The official 5.4 paths that returned HTTP 404 are listed in Sources. They are absence checks, not product pages.
![]()
Official Z.ai release notes (docs.z.ai/release-notes/new-released) as of 23 Nov 2026, Asia/Shanghai (HTTP 200). The latest model entries are GLM-5.3-Flash (26 Aug 2026) and GLM-5.3 (18 Aug 2026). The same page continues to GLM-5.2 (16 Jun 2026). There is no GLM-5.4 note.
That crop is the English changelog. The latest model rows are GLM-5.3-Flash dated 2026-08-26 and GLM-5.3 dated 2026-08-18. A missing row is the evidence. It is not a slip, and it is not a scheduled date. Further down the same page, still with no 5.4 row, the log lists GLM-5.1 on 2026-04-07 and GLM-5 on 2026-02-12.
![]()
Official BigModel 发布记录 (docs.bigmodel.cn/cn/update/new-releases) as of 23 Nov 2026, Asia/Shanghai (HTTP 200). The latest model entries are GLM-5.3-Flash (26 Aug 2026) and GLM-5.3 (dated 2026-8-19 in BigModel typography). GLM-5.2 is 2026-06-16. There is no GLM-5.4 note.
BigModel dates GLM-5.3 as 2026-8-19, without a leading zero. That is the vendor’s own typography for the same GLM-5.3 entry, not a second launch. Both changelogs agree on GLM-5.3-Flash: 2026-08-26. The BigModel page also lists GLM-5 on 2026-02-12 and GLM-5.1 on 2026-04-07, below the crop. Neither page has GLM-5.4 or GLM-5.5.
What Z.ai and Zhipu have officially confirmed
Current flagship and Flash sibling
Names and vendor IDs below are from the Z.ai model docs checked 23 Nov 2026. SandBase IDs for the live models add a z-ai/ prefix. That prefix does not create an official Z.ai SKU for GLM-5.4.
| Product | Vendor API ID | What the official docs say |
|---|---|---|
| GLM-5.3 | glm-5.3 | Latest flagship. Text only. 1M context. Max output 128K. Always-on reasoning: low, high, or max (default max). Same base model as GLM-5.2; the gains are from post-training. |
| GLM-5.3-Flash | glm-5.3-flash | First native multimodal model in the GLM-5 series. 320B total / 18B active mixture-of-experts. Sparse and linear attention. 1M context, max output 128K. Official copy calls it the first open-source frontier model to combine those two attention types. |
| GLM-5.3-FlashX | glm-5.3-flashx | High-speed tier. The Flash docs say it is live and delivers inference speeds of 200 tokens/s. Shares Flash’s multimodal design. Model codes on that page also include glm-5.3-flashx-legal. |
| GLM-5.3-FlashX-Legal | glm-5.3-flashx-legal | On official Z.ai pricing and the Flash docs. Not a SandBase card in this check, and not a call to action here. |
Official docs say max output is 128K for GLM-5.3 and for GLM-5.3-Flash. SandBase live cards print 131.1K. Both numbers stay attached to the page that printed them. This note does not collapse them into one figure.
The Flash docs sentence is “delivering inference speeds of 200 tokens/s”. The SandBase FlashX card description says “up to 200 tokens per second”. Those stay attached to the page that printed them.
Z.ai’s GLM-5.3 docs claim a 50% coding gain over GLM-5.2 on Z.ai Code Bench, and leading open-source results on Terminal Bench 3.0. Those are vendor claims. How to read them is already written up in GLM-5.3 Release Date & API Pricing (2026): Review, Benchmarks. The release notes also include a vendor-reported cybersecurity claim about vulnerabilities found in Z.ai’s own testing. This page does not walk through that testing. The comparison that already covers the claim is GLM-5.3 vs Mythos 5 vs Daybreak: Cybersecurity.
Flash pages this status note does not rewrite: GLM-5.3-Flash Multimodal and 1M Context, GLM-5.3-Flash Open-Weight Deployment, GLM-5.3-Flash Coding Benchmarks Explained, GLM-5.3-Flash Pricing: Agent Cost Guide, and GLM-5.3-Flash vs DeepSeek: Agent Cost Comparison. The earlier open-weight coding note in the line is GLM-5.1: The Open-Weight Model That Tops SWE-bench Pro.
Official changelog dates
The point releases that do exist are on the official logs. The guess that the next point must be named GLM-5.4 is not.
| Changelog | GLM-5 | GLM-5.1 | GLM-5.2 | GLM-5.3 | GLM-5.3-Flash | GLM-5.4 |
|---|---|---|---|---|---|---|
| Z.ai new releases | 2026-02-12 | 2026-04-07 | 2026-06-16 | 2026-08-18 | 2026-08-26 | No entry |
| BigModel release log | 2026-02-12 | 2026-04-07 | 2026-06-16 | 2026-8-19 | 2026-08-26 | No entry |
The page title on the BigModel log is 模型与产品发布记录. The date format on the GLM-5.3 row is the vendor’s, with no leading zero on the month. No point number between 5 and 5.3 is missing from that log. A complete past sequence is not a release date for a name that is not on the log.
Official prices
Three tariffs, kept apart. Z.ai publishes US dollars. BigModel publishes Chinese yuan. SandBase live cards are a third set of numbers, and two of them are discounted against the Z.ai dollar list. A yuan cell is not a dollar cell.
Z.ai official USD, per 1M tokens, from the pricing page checked 23 Nov 2026. Also linked from z.ai/pricing. Official typography uses $1.4 and $4.4 for GLM-5.3. SandBase prints that same list as $1.40 and $4.40. There is no GLM-5.4 row.
| Model | Input | Cached input | Cached input storage | Output |
|---|---|---|---|---|
| GLM-5.3 | $1.4 | $0.26 | Limited-time free | $4.4 |
| GLM-5.3-Flash | $0.15 | $0.03 | Limited-time free | $0.50 |
| GLM-5.3-FlashX | $0.37 | $0.075 | Limited-time free | $1.25 |
BigModel official CNY, per million tokens. The page labels the unit 元. The figures below are that tariff, read from the visible 5.3, Flash, and FlashX rows on 23 Nov 2026. They are not converted into the USD table. There is no GLM-5.4 row.
| Model | Context | Input | Output | Cache storage | Cache hit | Modalities |
|---|---|---|---|---|---|---|
| GLM-5.3 | 1M | 8 | 28 | Limited-time free | 2 | Text |
| GLM-5.3-Flash | 1M | 0.8 | 2.8 | Limited-time free | 0.23 | Image, video, file, text |
| GLM-5.3-FlashX | 1M | 2 | 7 | Limited-time free | 0.57 | Image, video, file, text |
No conversion is applied here.
SandBase prices are in the section on what you can call. They are not these official rows. On SandBase, GLM-5.3 and GLM-5.3-Flash show 25% and 40% off the Z.ai USD list. GLM-5.3-FlashX on the SandBase card matches the official Z.ai input and output ($0.37 / $1.25) and states no discount. The official FlashX cached-input price is $0.075. The SandBase HTML hero prints a separate cache-read price of $0.09, with no struck list beside it. This note does not turn $0.09 and $0.075 into a discount relationship. The card does not state one.
Where the name “GLM 5.4” comes from
Confirmed facts are above. This section is what is still absent from the official pages, or what showed up only in secondary writeups and investor comments.
The official logs run GLM-5, GLM-5.1, GLM-5.2, GLM-5.3, then GLM-5.3-Flash, with no skipped point number. That unbroken sequence is why people search “GLM 5.4”. The next-name guess is not a changelog entry. GLM-5.3 reused the GLM-5.2 base, and the docs attribute the gains to post-training. That history does not date a 5.4.
Third-party pages already target the query. They are not Z.ai or BigModel, and this check does not use them for an official name, ID, date, or price. Live on 23 Nov 2026: GLM 5.4: Does the Model Exist, and What Z.ai Ships Today; a Chinese comparison titled Qwen4-Max 对比 GLM-5.4:一个已发布,一个错失窗口期 (orcarouter.ai); a KoalaAPI note dated 2026-11-04, GLM-5.4 preview and agent coding, which still treats the model as unannounced; and Modelcast’s GLM 5.4 forecast. Modelcast’s own words are “this is an estimate, not news”, for a late-November to late-December 2026 window, last checked on that page 22 Nov 2026. Its table dates GLM-5 as 11 Feb 2026. The official Z.ai log says 2026-02-12. Use the official log.
An investor comment on Xueqiu, edited 22 Nov 2026 at 16:22 Asia/Shanghai, by 坚决的回报小巡游, claims GLM-5.4 is a transitional version post-trained from a 3T GLM-5.5, and that a larger GLM-5.5 is still coming in the fourth quarter, citing a 16 Nov conversation between JPMorgan and Zhipu management. Grade: Reported / Unknown. It is not a release note. This page does not say GLM-5.4 has shipped.
Union Alpha was rumored as a GLM test and later attributed to Pareto 26.9, associated with The Unbiased Company, with a Llama 3 tokenizer fingerprint. This check did not re-run that fingerprint. Union Alpha is not a GLM-5.4 signal.
Some secondary writeups say Tang Jie named GLM-6.0 on a 31 Aug 2026 results call. Grade: Reported / Unknown. GLM-6.0 is not in the release notes or the pricing tables. That claim does not cancel a possible intermediate point release, and it is not the official next name. No catalog path is linked, because this check found no live card.
A SandBase site search for glm-5.4 is fuzzy and can return other vendors’ “5.4” and “4.5” cards. A search hit is not a model card. The Z.ai hub lists GLM 5.2, GLM 5.3, GLM 5.3 Flash, GLM 5.3 FlashX, older GLM 4.5 Air, and a GLM 5.5 tile. It does not list GLM 5.4.
The same kind of note for other names is Is GPT-6.2 Out? and Is Qwen 4 Out?. Each has its own check date. Neither is evidence about GLM-5.4.
GLM-5.4 and GLM-5.5 are different unconfirmed names
SandBase has a coming-soon card for z-ai/glm-5.5 and no card for GLM-5.4. Neither string is an official Z.ai announcement. The 5.5 card is the one that exists, so it is documented here as a listing you should not call. The longer watch on that card is the GLM-5.5 status note from 18 Nov 2026. This page does not start a second 5.5 watch.
The public card and its .md twin were both HTTP 200 on 23 Nov 2026. The HTML title is “Z.ai GLM 5.5 API: Pricing & Specs | SandBase”. The model ID is z-ai/glm-5.5. published_at is null. Status and pricing both say Coming soon. The body says the model is not yet available on SandBase, and that as of early October 2026 Z.ai has not officially announced GLM 5.5.
The card text still says “as of early October 2026”. Its datePublished and dateModified are 2026-10-04T04:34:35.447Z, which is 12:34 Asia/Shanghai on 4 Oct 2026. That sentence is stale card copy. This article’s check is 23 Nov 2026, Asia/Shanghai.
The specifications block prints max output 131,072 tokens, input Text, output Text, and capabilities for chat, reasoning, structured output, and function calling. The hero prints 131.1K and context N/A. Those are catalog placeholders on an upcoming card. They are not official Z.ai capabilities for GLM-5.5, and they are not specs for GLM-5.4. The card also says requests to this model ID return 404 model_not_available. The .md includes a curl aimed at z-ai/glm-5.5. That example is not a working call. It is not copied here. Do not send z-ai/glm-5.5, and do not invent a z-ai/glm-5.4 string to put in its place.
![]()
SandBase coming-soon card at /models/z-ai/glm-5.5, captured 23 Nov 2026, Asia/Shanghai (HTTP 200). Status and pricing are Coming soon, context is N/A, and a placeholder max output of 131.1K / 131,072 tokens is printed. That is not an official Z.ai spec, and it is not a GLM-5.4 spec. The card still says “as of early October 2026”; this post’s check date is 23 Nov. The dollar amount in the top banner is the site-wide first top-up bonus, not a model price. This is not a live callable listing.
The banner in that frame reads “Get 25% extra credits on your first top-up, up to $1,000.” That is a top-up bonus, not a model tariff.
The Z.ai models hub is a different page. Its GLM 5.5 tile says “See model page”, not “Coming soon”, and the hub tiles for the live models show list-style prices rather than the discounted heroes on the individual cards. The GLM 5.3 tile shows $1.4 / $4.4, GLM 5.3 Flash shows $0.15 / $0.5, and GLM 5.3 FlashX shows $0.37 / $1.25. Use the cards below for the rate on a live request. Do not quote a hub tile as the SandBase charge.
What you can call today on SandBase
Rechecked 23 Nov 2026, Asia/Shanghai. HTML and .md were both HTTP 200. Quote the HTML hero. That is where cache read is printed. The .md Facts block omits cache on all three live cards, including FlashX. The 18 Nov GLM-5.5 status note did not have a FlashX cache-read figure. The FlashX hero shows one now: $0.09 per 1M tokens, with no struck list price and no discount badge.
| Card | SandBase ID | Price per 1M tokens | Context / max output on the card |
|---|---|---|---|
| GLM-5.3 | z-ai/glm-5.3 | 25% OFF. Input $1.05 (struck $1.40), output $3.30 (struck $4.40), cache read $0.19 (struck $0.26). HTML hero. | 1M context, up to 131.1K output. Text. Official docs say 128K max output. |
| GLM-5.3-Flash | z-ai/glm-5.3-flash | 40% OFF. Input $0.09 (struck $0.15), output $0.30 (struck $0.50), cache read $0.02 (struck $0.03). HTML hero. | 1M context, up to 131.1K output. Vision: text and image. Official docs say 128K max output. |
| GLM-5.3-FlashX | z-ai/glm-5.3-flashx | Input $0.37, output $1.25, cache read $0.09. No struck list price. No discount badge. The .md FAQ says discounts, when available, are shown in the pricing on the page. | 1M context, up to 131.1K output. Vision: text and image. Official docs say 200 tokens/s and 128K max output. |
![]()
Live SandBase GLM-5.3 card at /models/z-ai/glm-5.3, captured 23 Nov 2026, Asia/Shanghai (HTTP 200). The HTML hero shows 25% OFF: $1.05 input / $3.30 output / $0.19 cache read per 1M tokens, with struck list prices $1.40 / $4.40 / $0.26. Context is 1M with 131.1K max output on the card. The Benchmarks block under the hero is a SandBase catalog chart. It is not an official Z.ai score, and this note does not quote it.
![]()
Live SandBase GLM-5.3-Flash card at /models/z-ai/glm-5.3-flash, captured 23 Nov 2026, Asia/Shanghai (HTTP 200). The HTML hero shows 40% OFF: $0.09 input / $0.30 output / $0.02 cache read per 1M tokens, with struck list prices $0.15 / $0.50 / $0.03. Context is 1M with 131.1K max output on the card. Input on the card is text and image.
![]()
Live SandBase GLM-5.3-FlashX card at /models/z-ai/glm-5.3-flashx, captured 23 Nov 2026, Asia/Shanghai (HTTP 200). The HTML hero shows $0.37 input / $1.25 output / $0.09 cache read per 1M tokens. There is no struck list price and no discount badge. Context is 1M with 131.1K max output on the card. Official Z.ai cached input for FlashX remains $0.075 on the Z.ai pricing page. The two cache figures are not combined into a discount.
The struck list prices on GLM-5.3 and GLM-5.3-Flash match the Z.ai USD list, printed with two decimals ($1.40 / $4.40 / $0.26 and $0.15 / $0.50 / $0.03). The live charge on those two cards is the discounted hero. It is not the struck list, and it is not the BigModel yuan tariff. The purple strip on each card is the same first top-up bonus. It is not a fourth price.
Which ID to put in the request:
- Text, and you want the current official flagship: GLM-5.3. Always-on reasoning. SandBase bills the 25% off hero.
- The call needs images: GLM-5.3-Flash. Official docs describe native multimodal input, including video, image, text, and file. The SandBase hero shows text and image, plus vision. That official modality list is not silently copied onto the SandBase card.
- You want the documented 200 tokens/s tier: GLM-5.3-FlashX, at the hero price of $0.37 / $1.25 / cache read $0.09, with no discount stated on that card.
GLM-5.2 is also live (z-ai/glm-5.2 at $1.26 / $3.96, 10% off list $1.40 / $4.40, 1M context, up to 32.8K output) at the GLM-5.2 card. It is the previous flagship. It does not stand in for a GLM-5.4 ID. New work goes to GLM-5.3 or Flash unless you already pin an older card.
How to call the live models
Language models use POST https://api.sandbase.ai/v1/chat/completions with Authorization: Bearer $SANDBASE_API_KEY. One SandBase API key is enough. A separate Z.ai signup is not required for these three cards. Do not send these IDs to POST /v1/run. That path is for media.
No paid run was made for this status note. The example uses z-ai/glm-5.3 because that card is live. Swap the model string for z-ai/glm-5.3-flash or z-ai/glm-5.3-flashx when that is the card you mean. Do not put z-ai/glm-5.4 or z-ai/glm-5.5 in this request.
curl -X POST https://api.sandbase.ai/v1/chat/completions \
-H "Authorization: Bearer $SANDBASE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "z-ai/glm-5.3",
"messages": [
{"role": "user", "content": "Hello"}
],
"max_tokens": 2048
}'Sign in is where /signup currently redirects. The chat-completions walkthrough is First API call. Account pricing sits on the SandBase pricing page. Open GLM-5.3, GLM-5.3-Flash, or GLM-5.3-FlashX before you send the request.
What we will check if a next official GLM is announced
A missing card, a coming-soon card, and a forecast page are not the announcement. The checks, in order, whether the official name is GLM-5.4 or something else:
- An official Z.ai or BigModel blog, docs page, or changelog entry that names the model.
- An official API ID and a price row. Weights, if Z.ai publishes them.
- After those exist, a fresh read of the SandBase HTML card and the matching
.md. A price on a card that is actually callable is the catalog test. A guessed path is not.
Until that happens, production traffic stays on GLM-5.3 for text, and on GLM-5.3-Flash or GLM-5.3-FlashX when the call needs the multimodal tier.
FAQ
Is GLM-5.4 out?
No. As of 23 Nov 2026, Asia/Shanghai, Z.ai and Zhipu have not officially released or announced GLM-5.4. Official release notes still end at GLM-5.3 (18 Aug 2026 on Z.ai, 2026-8-19 on BigModel) and GLM-5.3-Flash (26 Aug 2026 on both).
Can I call GLM-5.4 on SandBase today?
No. There is no SandBase card for GLM-5.4. The catalog paths checked on 23 Nov 2026 returned HTTP 404. Do not send a guessed z-ai/glm-5.4 id. z-ai/glm-5.5 is a different string, on a public coming-soon card, and the card says calls return 404 model_not_available.
What can I call today?
GLM-5.3 (z-ai/glm-5.3) at $1.05 / $3.30 / cache read $0.19 per 1M tokens, which is 25% off the struck list $1.40 / $4.40 / $0.26. GLM-5.3-Flash (z-ai/glm-5.3-flash) at $0.09 / $0.30 / cache read $0.02, 40% off $0.15 / $0.50 / $0.03. GLM-5.3-FlashX (z-ai/glm-5.3-flashx) at $0.37 / $1.25 / cache read $0.09, with no discount stated. Endpoint: POST https://api.sandbase.ai/v1/chat/completions.
Is the late-November window a release date?
No. A third-party forecast page estimates late November to late December 2026 and says that window is not news. It is not on the official Z.ai or BigModel pages checked 23 Nov 2026.
Did an investor post confirm that GLM-5.4 already shipped?
No. A Xueqiu comment edited 22 Nov 2026 discusses a transitional 5.4 and a later 5.5. That is investor discussion. It is not an official announcement, and it is not used here as a ship date, an API ID, or a price.
Did Z.ai announce GLM-6.0 instead?
Not in this check. GLM-6.0 does not appear in the release notes or the pricing tables. It is not treated as the official next name, and it does not by itself rule out a later point release that Z.ai has not announced.
Sources
Confirmed on 23 November 2026, Asia/Shanghai:
- Z.ai release notes. Latest model entries: GLM-5.3-Flash on 2026-08-26, GLM-5.3 on 2026-08-18. Earlier rows include GLM-5.2 on 2026-06-16, GLM-5.1 on 2026-04-07, and GLM-5 on 2026-02-12. HTTP 200. No GLM-5.4 row.
- BigModel release log. GLM-5.3-Flash on 2026-08-26, GLM-5.3 dated 2026-8-19, GLM-5 on 2026-02-12. HTTP 200. No GLM-5.4 row.
- Z.ai GLM-5.3 docs and GLM-5.3-Flash / FlashX docs.
- Z.ai pricing, USD and BigModel pricing, CNY. No GLM-5.4 row on either table.
- Z.ai llms.txt. No GLM-5.4 page and no GLM-5.5 page.
- chat.z.ai. The title still names GLM-5.3-Flash.
- Hugging Face model search
https://huggingface.co/api/models?search=glm-5.4&limit=10returned an empty list. - OpenRouter
https://openrouter.ai/api/v1/models. No glm-5.4 id and no glm-5.5 id. Present z-ai GLM ids include glm-5, glm-5-turbo, glm-5.1, glm-5.2, glm-5.3, glm-5.3-flash, glm-5.3-flashx, glm-5.3-prime, and glm-5v-turbo, plus older 4.x ids. - Official 5.4 paths that returned HTTP 404:
https://z.ai/blog/glm-5.4,https://docs.z.ai/guides/overview/glm-5.4,https://docs.z.ai/guides/llm/glm-5.4,https://docs.z.ai/guides/vlm/glm-5.4, andhttps://docs.bigmodel.cn/cn/guide/models/text/glm-5.4. - Live SandBase cards: GLM-5.3, GLM-5.3-Flash, GLM-5.3-FlashX, GLM-5.2. Coming-soon listing, not callable: GLM-5.5. Hub: Z.ai models. Catalog paths for a GLM-5.4 card returned HTTP 404 and are not linked as product pages.
Reported / Unknown only, and not used for official names, IDs, or prices: third-party forecast and comparison pages that already target the query, the Modelcast late-November to late-December 2026 estimate (the page says it is not news), the 22 Nov 2026 Xueqiu investor comment, the retired Union Alpha rumor, and a secondary claim that GLM-6.0 was named on a 31 Aug 2026 results call.
Sources checked 23 November 2026, Asia/Shanghai. Official docs for names, dates, and prices. SandBase cards for what you can call. Rumors stay in the unconfirmed section.
Found this useful? Share it