Model ID and API Format
Z.ai's official announcement uses the request model ID glm-5.3, with always-on thinking and a new reasoning_effort parameter (low/high/max).
GLM-5.3 API
Z.ai / Zhipu AI-Text-only · released Aug 14, 2026 · API staged-Released Aug 14, 2026 · Per-token API staged · EvoLink route pending
GLM-5.3 launched on August 14, 2026 — but its per-token API is rolling out in stages with no published pricing, so the practical question is when you can actually call it. Join the alert to be notified when EvoLink verifies a callable route.
Is the GLM-5.3 API Available Now?
Official Release and API Status
Released Aug 14, 2026; per-token API staged, pricing unpublished
EvoLink Access Status
Route not yet verified; launch alerts are open
Last Verified
August 14, 2026
EvoLink Availability Watch
Get Notified When the GLM-5.3 Route Is Verified Live
We will email you when EvoLink verifies a callable GLM-5.3 route, the request model ID, live pricing, and supported API behavior.
Join the GLM-5.3 Launch Alert
We will only email you about GLM-5.3 availability. No spam.
Want faster updates? Join the EvoLink Discord
What You Will Receive When Access Opens
When access opens we will email you a launch alert, and this page will be updated with the verified model ID, live pricing, API compatibility, regions and limits — plus the path to integrate through EvoLink's unified API while keeping a fallback model.
What the Release Confirmed — and What It Left Open
The August 14 release settled some integration fields and left others open. EvoLink updates each one only after official publication and route-level verification.
Input and Output Pricing
Per-token pricing is unpublished — Z.ai's own pricing page still stops at GLM-5.2. The live EvoLink pricing module will be the source of truth after launch.
Context Window and Maximum Output
Official documentation lists a 1M-token context window and up to 128K output tokens for GLM-5.3.
Regions, Concurrency, and Rate Limits
Regions, concurrency, quotas, rate limits, and 429 behavior require a callable production route; none are published yet.
Currently Available API Providers
On release day no aggregator could serve GLM-5.3, and the BigModel API page says "coming soon." Open weights are promised about two weeks after launch.
What EvoLink Will Verify Before Recommending Production Use
Only real workloads on a real API route can answer the production question. These six user-demand metrics are what EvoLink will verify after access opens; vendor benchmarks are not a substitute.
Coding Task Completion Rate
Measure whether bug fixes, multi-file refactors, and test tasks actually pass instead of judging one response.
Tool-Call Success and Agent Loops
Track argument errors, repeated calls, unproductive loops, and how often a person must take over.
Effective Long-Context Reliability
Test whether the model keeps constraints, finds evidence, and completes work deep inside large repositories and documents.
Always-On Thinking Cost
GLM-5.3 removes the option to disable thinking. Measure latency and output-token overhead at each reasoning_effort level against your former non-thinking paths.
Latency, Concurrency, and Reliability
Evaluate time to first token, end-to-end speed, capacity, error rate, and peak-hour 429 responses.
Cost per Accepted Task
Include token use, retries, failed runs, and human correction instead of comparing list prices alone.
While the API Is Staged: Ship on GLM-5.2
GLM-5.3 is not yet callable by the token, and its pricing is unpublished — so teams should not delay a project that can ship on GLM-5.2 today.
Should You Use GLM-5.2 Now?
If delivery matters now, establish a real baseline on the documented and callable GLM-5.2 route.
When Is GLM-5.3 Worth Switching To?
Switch only when your workloads show a verified improvement on a real route — and note the breaking change: thinking can no longer be disabled.
How Do You Keep Switching and Fallback Simple?
Keep the current model behind EvoLink's unified API, then add GLM-5.3 after validation without rebuilding another provider integration.
Related Model Listings on EvoLink
While the GLM-5.3 API is staged, these are the routes and watches most relevant to the same decision.
GLM-5.2
LiveThe callable member of the GLM family today — 1M context, published pricing, live EvoLink route. The practical baseline while GLM-5.3's API is staged.
View model
GLM 5.5
WatchA separate, still-unannounced model — possibly the release after GLM-5.3 — tracked on its own availability watch.
View model
Claude Opus 4.8
LiveA live Claude route many production coding agents run on today; the cross-vendor comparison target for GLM-5.3 evaluations.
View modelLearn More About GLM-5.3
The release is staged: what shipped, what's pending, and whether to switch.
GLM-5.3 release tracker
What shipped on August 14, what's staged, and the open-weight timeline — updated as status changes.
GLM-5.3 vs GLM-5.2
What actually changed, the always-on-thinking breaking change, and when to stay on GLM-5.2.
GLM 5.5 availability watch
A separate, still-unannounced model — possibly the release after GLM-5.3 — tracked on its own page.
GLM-5.3 API FAQ
Is the GLM-5.3 API available now?
Not by the token. On release day, GLM-5.3 was available only through the GLM Coding Plan subscription and ZCode; Z.ai's pricing page has no glm-5.3 entry and the BigModel API page says "coming soon."
What is the GLM-5.3 model ID?
Z.ai's official announcement uses glm-5.3 in its API example. Verify against official API documentation before production use.
How much will the GLM-5.3 API cost?
Unpublished. Z.ai's pricing page still lists only GLM-5.2 ($1.40/$0.26/$4.40 per million tokens — GLM-5.2's own rates). This page will use EvoLink's live pricing module only after the route and SKUs are approved.
Is GLM-5.3 available on EvoLink?
Not yet verified. EvoLink publishes a callable route, model ID, and pricing only after real request, billing, and behavior checks pass. Join the alert to be notified.
What changed compared with GLM-5.2?
Same base model — all gains come from post-training. The confirmed breaking change: thinking can no longer be disabled, and reasoning_effort (low/high/max) becomes the control surface.
Does GLM-5.3 support images, screenshots, or PDFs?
No. GLM-5.3 is text-only; Z.ai's multimodal work remains in the separate GLM-V line.
When will the open weights be released?
Z.ai promises the weights about two weeks after launch (around August 28, 2026), after a safety evaluation. The license has not been stated.
Is GLM-5.3 the same thing as GLM 5.5?
No — they are separate models. GLM 5.5 remains unannounced; if it comes, it is expected as a later release, and it is tracked on its own availability page.
What should I use while the API is staged?
GLM-5.2 — live, priced, 1M context, callable through EvoLink today. Keep the model ID configurable so switching to a verified GLM-5.3 route is a configuration change.
What does the EvoLink alert provide?
A notification after EvoLink verifies the GLM-5.3 route: official release status, model ID, pricing, route behavior, and access state.