People searching glm5.3-flashx on sub plan, plus the Chinese questions about Zcode, OpenCode, and "not enabled yet," are stuck on one product split. GLM-5.3-Flash and GLM-5.3-FlashX share a documentation page, but they do not share an entitlement.
This page was checked against the Z.ai GLM-5.3-Flash/FlashX overview, the Z.ai pricing table, and the published GLM-5.3-Flash usage-campaign notice on October 12, 2026. We did not sign into a Coding Plan account for this article. If a dashboard disagrees with the docs, trust the live docs page and re-check the date, because this is the sentence most likely to change.
Short answer: the API model code glm-5.3-flashx is documented and priced. The same official overview says GLM-5.3-FlashX is not yet available on the GLM Coding Plan. Zcode and OpenCode access that is written down today is for GLM-5.3-Flash, not for the FlashX tier.
Two doors, not one toggle
Z.ai's overview lists both model codes on one page: glm-5.3-flash and glm-5.3-flashx. Text settings match GLM-5.3, with a 1M-token context window and a 128K maximum output. Recommended sampling on that page is temperature: 1, top_p: 0.95, and reasoning_effort: max. thinking.type only supports enabled.
The product difference is serving speed, not a second set of weights. The overview says FlashX is live and quotes inference speeds of 200 tokens/s. Treat that as the vendor's ceiling claim. It is not an independent median we measured for this article.
API access: the path that is actually open
On the Z.ai pricing page checked the same day, prices are USD per 1 million tokens:
| Model | Input | Cached input | Output |
|---|---|---|---|
| GLM-5.3-Flash | $0.15 | $0.03 | $0.50 |
| GLM-5.3-FlashX | $0.37 | $0.075 | $1.25 |
| GLM-5.3 | $1.40 | $0.26 | $4.40 |
FlashX is 2.5 times Flash on input, cached input, and output ($0.37 / $0.15, $0.075 / $0.03, and $1.25 / $0.50). Cached-input storage is listed as limited-time free for these rows. Third-party routers can show a different cache rate than this table. Use the Z.ai table when you are calling Z.ai directly, and re-check it before you budget a month of agent traffic.
A 1M-token cache-miss input plus 100K output tokens is $0.37 + $0.125 = $0.495 at the FlashX list rates above, versus $0.15 + $0.05 = $0.20 on standard Flash. That gap is the whole subscription argument: the fast tier is an API upsell, not a free switch inside a plan.
Call the model with the code glm-5.3-flashx on Z.ai's chat-completions API. Image input, on the same docs page, is an image_url content block (URL or base64). The page also says the model accepts video and files as visual input.
What the subscription plan includes today
The GLM Coding Plan section of the overview is explicit, and it is easy to miss because it sits in parentheses:
GLM-5.3-Flash can be used with your preferred tools, with 3× the available quota compared with GLM-5.3. (GLM-5.3-FlashX is not yet available on the plan.)
The plan uses points, not the per-token table above. Off-peak hours, including all day on weekends, consume 50% of the standard points. That discount is written for plan usage in general. It does not put FlashX onto the plan.
So the search glm5.3-flashx on sub plan has a documented answer as of October 12, 2026: not yet. A social post or an application form is not the same thing as the docs line. We are not treating those as access. If Z.ai later opens FlashX on the plan, this paragraph is the one to update.
If you want GLM 5.3 Flash in a browser without sorting plan quotas, open it on glm5.app. The model id documented for that site is glm-5.3-flash, which is the standard Flash tier, not glm-5.3-flashx.
Zcode and OpenCode
The questions "is it in Zcode yet?" and "does OpenCode have it?" are the same entitlement problem with a tool name attached.
Z.ai published a GLM-5.3-Flash usage campaign for paid Coding Plan users, running September 3, 2026 through October 7, 2026 (Singapore time, UTC+8). During 23:00–09:00, Zcode and AutoClaw usage of GLM-5.3-Flash was zero-quota, and other supported agents had doubled quota. The notice says the campaign applies only to GLM-5.3-Flash, and that Zcode support starts at version 3.10. Today is October 12, 2026, which is after the published end date. Do not plan on that window still being open, and do not read it as FlashX access. It never said FlashX.
OpenCode shows up because Z.ai tested GLM-5.3-Flash anonymously as ox-alpha on OpenCode and OpenRouter before the public launch, then rolled GLM-5.3-Flash out to Coding Plan users. That history explains why people ask whether OpenCode "already has FlashX." The current overview does not list FlashX as a plan model, so an OpenCode picker that only shows plan models should not be expected to offer glm-5.3-flashx.
If your account says it is not enabled
- Confirm which product you are in. A Coding Plan seat follows the parenthetical on the overview: Flash yes, FlashX no, until that sentence changes.
- If you need the fast tier now, use the metered API and the model code
glm-5.3-flashx, and budget about 2.5× the Flash token price from the table above. - If you need the plan's 3× Flash quota, stay on
glm-5.3-flash. On the Z.ai table checked October 12, 2026, that tier is $0.15 input, $0.03 cached input, and $0.50 output per 1M tokens. - Ignore typos and alternate spellings (
flashxvsflash xvs a missing hyphen). There is one API code:glm-5.3-flashx.
FAQ
Is GLM-5.3-FlashX included when I subscribe?
Not on the evidence of the official overview checked October 12, 2026. The plan includes GLM-5.3-Flash at 3× the GLM-5.3 quota. FlashX is called out as not yet available on the plan.
Can I apply for FlashX on the Coding Plan?
We did not find that application on the official overview or pricing page. Until the docs drop the "not yet available" line, assume API access is the supported route.
Does Zcode or OpenCode already have GLM 5.3 FlashX?
Documented tool access is for GLM-5.3-Flash. The Zcode night-time campaign was Flash-only and its published end date was October 7, 2026. OpenCode was a pre-release test surface for Flash (ox-alpha), which is a different model code from glm-5.3-flashx.
What context and output limits should I set?
The overview lists a 1M context window and 128K maximum output for this Flash/FlashX page. It also says thinking cannot be disabled (thinking.type is enabled only).
Sources
- GLM-5.3-Flash/FlashX overview — model codes, 1M / 128K limits, sampling defaults, and the Coding Plan line that FlashX is not yet on the plan. Checked October 12, 2026.
- Z.ai pricing — list rates for Flash, FlashX, and GLM-5.3. Prices move; re-check before you invoice a workload.
- GLM-5.3-Flash usage campaign — Zcode / other-agent quota rules for Flash only, with a published end date of October 7, 2026.
- GLM-5.3-Flash announcement — Coding Plan rollout and the
ox-alphaOpenCode / OpenRouter test, for Flash rather than FlashX.
