01 · Precedent
5.3 had variants
GLM-5.3 shipped as GLM-5.3-Flash and GLM-5.3-FlashX. A Flash-first launch of the successor would follow the same pattern.
Pre-release tracker · Z.ai / Zhipu AI
GLM 5.5 has not been released yet. This page tracks what is confirmed, what is only reported, and which GLM models you can call on TokenRa today.
Z.ai has published no model card, API pricing, or benchmark table for Zhipu's next flagship. Everything below is either an official confirmed fact or an unconfirmed report. We do not fill the gaps with estimates.
No official specs exist for Zhipu's next flagship, so this table separates confirmed GLM-5.3 numbers from what has been reported—not confirmed—about the successor.
| Dimension | GLM-5.3-Flash (confirmed) | Successor to GLM 5.3 (reported / unconfirmed) |
|---|---|---|
| Parameters | 320B total / 18B active | 2.5T–3.5T circulates online; no official source |
| Context window | 1,310,720 tokens (~1M) | Not announced |
| Max output | 131,072 tokens | Not announced |
| Input modalities | Text, image, video | Not announced |
| Reasoning | Always on; reasoning_effort sets depth | Not announced |
| License | Proprietary API access | Not announced |
| API pricing | $0.14 / $0.52 / $0.04 per 1M tokens (in/out/cache) | Not announced |
| Availability on TokenRa | Live now | Only after Z.ai releases the model |
Everything in this table is a source and a date. Anything marked unconfirmed stays unconfirmed until Z.ai says otherwise.
| Date | What happened | Status |
|---|---|---|
| 2026-06-30 | CGTN reports Zhipu's next flagship is expected in August 2026. | Reported |
| 2026-08 | August passes with no release and no model card from Z.ai. | Confirmed |
| 2026 (undated) | A glm-5.5-flash identifier appears in third-party tooling and model listings, discussed on Reddit. | Reported |
| 2026-11 | A new round of release chatter. No official date or page from Z.ai. | Reported |
| 2026-11-14 | SERP for "glm 5.5" is still dominated by speculation. Z.ai publishes GLM-5 material only. | Confirmed |
The GLM 5.3 generation shipped in two speeds — Flash and FlashX — so a smaller variant of the successor appearing before or alongside the full model is a reasonable expectation, not a confirmed plan.
01 · Precedent
GLM-5.3 shipped as GLM-5.3-Flash and GLM-5.3-FlashX. A Flash-first launch of the successor would follow the same pattern.
02 · Reported trace
A glm-5.5-flash identifier has been reported in third-party model listings. Z.ai has not confirmed it.
03 · Available now
1M-token context, native multimodal, $0.14 / $0.52 per 1M input/output tokens on TokenRa. See pricing →
We will not open a separate URL for it. The model ID, price, and limits will be added to this section on release day.
The successor to GLM 5.3 is not callable yet. The GLM-5.3 family is live, and it is the closest available option until the new model ships.
| Model | Context | Input / 1M | Output / 1M | Cache read / 1M |
|---|---|---|---|---|
| GLM-5.3-Flash | 1,310,720 tokens | $0.14 | $0.52 | $0.04 |
| GLM-5.3-FlashX | 1,310,720 tokens | $0.14 | $0.52 | $0.04 |
| GLM-5.3 | See docs | See docs | See docs | See docs |
Docs: GLM-5.3 Flash parameters · GLM-5.3 FlashX parameters
GLM-5.3-Flash and FlashX share the same pricing. FlashX delivers up to 200 tok/s output speed — useful for latency-sensitive applications — while Flash suits workloads where throughput matters more than speed. Both accept text, image, and video input and use the same OpenAI-compatible endpoint on TokenRa. Confirm current limits and discounts in the console before production use.
Same URL, rewritten in place. In order:
Step 1
The status block above changes to "Released", with Z.ai's announcement date and a source link.
Step 2
Official context window, modalities, license, benchmarks, and API price replace every reported figure.
Step 3
The model ID, availability date on TokenRa, and a working request example are added — nothing is published before it actually works.
No. As of 2026-11-15 there is no official announcement, model card, or API pricing page from Z.ai. This page is a pre-release tracker and is updated as information is confirmed.
It is possible but unconfirmed. The GLM 5.3 generation shipped as Flash and FlashX variants, so a smaller variant appearing before the full model is plausible. A glm-5.5-flash identifier was reported in third-party model listings, but Z.ai has not confirmed it. If a smaller variant ships first, this page will cover it under the same URL.
No official release date has been announced. A June 2026 report said August 2026, and that window passed without a release. Treat any date that does not come directly from Z.ai as unconfirmed.
API pricing has not been announced. GLM pricing is published with the official model card. As a reference, GLM-5.3-Flash costs $0.14 per 1M input tokens, $0.52 per 1M output tokens, and $0.04 per 1M cached read tokens on TokenRa.
Not yet. Access can only open after Z.ai releases the model and it is enabled on TokenRa. Until then the GLM-5.3 family is available: GLM-5.3, GLM-5.3-Flash, and GLM-5.3-FlashX.
Unknown. No official benchmarks have been published. Parameter counts circulating online are unconfirmed and should not be treated as specifications until Z.ai publishes a model card.
The license has not been announced. Previous GLM releases have included open-weight versions, but no license statement exists for the successor yet.
Not announced. As a reference, GLM-5.3-Flash offers a 1,310,720-token context window with native multimodal input across text, image, and video.