Z.ai API errors

Z.ai API error codes, and how to fix each one

All 35 codes Z.ai's GLM API returns, from 1113 to the Coding Plan limits, in plain words: what each one means, whether retrying helps, and what to change. Checked against Z.ai's own error table on October 22, 2026.

Independent reference, not affiliated with Z.aiUpdated for the October 16 API key split

The six you are most likely seeing

1113Fix, then resend

Call billed to an empty balance

Use a Coding Plan key at a coding endpoint, or top up the balance.

Read the fix

1302Retry with backoff

Your account's rate limit

Send fewer requests at once and retry with backoff.

Read the fix

1305Retry with backoff

Z.ai is overloaded

Retry with exponential backoff, or call the model through another host.

Read the fix

1308Wait for the reset

Plan allowance used up

Wait for the reset time in the message; off-peak calls cost half the credits.

Read the fix

1311Fix, then resend

Model not in your Coding Plan

Set every model in your tool to glm-5.3 or glm-5.3-flash.

Read the fix

1261Fix, then resend

Input longer than the context window

Trim the history (/compact in Claude Code) or use a 1M-context model.

Read the fix

GLM Coding Plan

Four things a plan call needs

Most Coding Plan errors (1113, 1311, a balance that drops while plan credits sit unused) come from one of these four being wrong. Z.ai bills a call to the plan only when all of them hold; otherwise it bills your pay-as-you-go balance or refuses the call.

Z.ai API key notice, Oct 16, 2026 Z.ai API Keys page

The full GLM Coding Plan guide, with setup

The key: A key from API Keys, Personal Coding Plan (Team members: Team Coding Plan, My Plan). Since October 16, 2026 a Standard API Key never draws on the plan.
The endpoint: https://api.z.ai/api/anthropic for Claude Code, https://api.z.ai/api/coding/paas/v4 for OpenAI-compatible tools. Not https://api.z.ai/api/paas/v4, which bills the balance.
The model: glm-5.3 or glm-5.3-flash, the only two the plan calls (older ids are rerouted to them).
The tool: One of Z.ai's supported coding tools. Your own scripts and apps need a Standard API Key.

Reading an error

Two codes in every error

Each failure carries an HTTP status and, inside the JSON body, Z.ai's own business code. The business code is the one that says what went wrong: a 429 can be a rate limit, an empty balance or a spent plan.

HTTP/2 429
{"error":{"code":"1113","message":"Insufficient balance or no resource package. Please recharge."}}

Coding tools show the same JSON inside their own error line, so look for the four-digit number after code.
Z.ai's example sends the code as a string ("1113") though its schema calls it an integer, so compare it as a string in your own error handling.
OpenAI's and Anthropic's SDKs retry a 429 twice by default, and coding agents retry too, so a 1113 or a plan limit can look like a hang before it surfaces.
Once a response is streaming, failures arrive as a finish_reason instead (see below).

What to do

Retry, wait, or fix

Retry with exponential backoff and jitter

Rate limits, overload and Z.ai's own server errors. For 1302, also send fewer requests at once.

1302 1305 1234 1200 1230 HTTP 500

Wait for the reset time in the message

Coding Plan limits. The message states when the allowance refills; retrying before then fails again.

1308 1310 1316 1317 1318 1319 1320 1321

Fix something, then resend

The key, the endpoint, the model, the plan or the request is wrong. These fail every time until it changes, including 1113 and 1311 although they arrive as HTTP 429.

1113 1311 1315 1220 1309 1313 1314 1211 1210 1214 1213 1215 1212 1222 1221 1261 1000 1001 1003 1005 1301

Every code

The full lookup table

Z.ai's complete list, in code order. Each code links to its explanation below.

Code	HTTP	What it means	What to do
1000	401	API key rejected	Fix, then resend
1001	401	No key in the request	Fix, then resend
1003	401	Token expired	Fix, then resend
1005	401	Account needs two-factor	Fix, then resend
1113	429	Call billed to an empty balance	Fix, then resend
1200	500	Server-side call error	Retry with backoff
1210	400	Invalid parameter	Fix, then resend
1211	400	Model id not recognised	Fix, then resend
1212	400	Model does not support this call	Fix, then resend
1213	400	Required field missing	Fix, then resend
1214	400	A field's value is not allowed	Fix, then resend
1215	400	Two fields that conflict	Fix, then resend
1220	403	No access to this API	Fix, then resend
1221	400	API or model retired	Fix, then resend
1222	400	Wrong path or base URL	Fix, then resend
1230	500	Server-side processing error	Retry with backoff
1234	500	Transient network error	Retry with backoff
1261	400	Input longer than the context window	Fix, then resend
1301	400	Content filter triggered	Fix, then resend
1302	429	Your account's rate limit	Retry with backoff
1305	429	Z.ai is overloaded	Retry with backoff
1308	429	Plan allowance used up	Wait for the reset
1309	429	Coding Plan expired	Fix, then resend
1310	429	Weekly or monthly limit used up	Wait for the reset
1311	429	Model not in your Coding Plan	Fix, then resend
1313	429	Throttled under the fair-use rules	Fix, then resend
1314	429	Enterprise package expired	Fix, then resend
1315	429	Enterprise coding key used elsewhere	Fix, then resend
1316	429	5-hour limit, no balance for overage	Wait for the reset
1317	429	7-day limit, no balance for overage	Wait for the reset
1318	429	5-hour limit, overage capped	Wait for the reset
1319	429	7-day limit, overage capped	Wait for the reset
1320	429	5-hour limit, overage capped	Wait for the reset
1321	429	7-day limit, overage capped	Wait for the reset
HTTP 500	500	Internal error (no business code)	Retry with backoff

Z.ai error table (docs.z.ai)

4 codes

Balance, key type and plan access

The errors behind most 'it worked yesterday' reports since Z.ai split its API keys on October 16, 2026. They arrive as HTTP 429 or 403, but none of them clears by retrying.

Error 1113

HTTP 429Fix, then resend

“Insufficient balance or no resource package. Please recharge.”

What it means

Z.ai billed the call to your pay-as-you-go balance and found nothing to charge. With a GLM Coding Plan it means the call did not qualify for the plan: since October 16, 2026 a Standard API Key always bills the balance, an older Compatible key bills the balance when it calls the general endpoint (https://api.z.ai/api/paas/v4), and the plan pays only for calls from Z.ai's supported coding tools. It arrives as HTTP 429, so SDKs and coding tools retry it before showing it, but it never clears by retrying.

How to fix it

With a plan: create a key under API Keys, Personal Coding Plan (Team members: Team Coding Plan, My Plan) and point the tool at a coding endpoint, https://api.z.ai/api/anthropic for Claude Code or https://api.z.ai/api/coding/paas/v4 for OpenAI-compatible tools. Without a plan: top up the balance, or call a model Z.ai lists as free (GLM-4.7-Flash and GLM-4.5-Flash).

Error 1311

HTTP 429Fix, then resend

“Your current subscription plan does not yet include access to ${model_name}”

What it means

The Coding Plan does not cover the model you asked for. Z.ai's FAQ says the plan calls only GLM-5.3 and GLM-5.3-Flash. It routes requests for GLM-5.2 and GLM-5.1 to GLM-5.3 and requests for GLM-4.7 to GLM-5.3-Flash, but anything else (a vision or image model, a Prime variant, an older generation) is outside the plan.

How to fix it

Set every model your tool uses to glm-5.3 or glm-5.3-flash. In Claude Code that means all three of ANTHROPIC_DEFAULT_OPUS_MODEL, ANTHROPIC_DEFAULT_SONNET_MODEL and ANTHROPIC_DEFAULT_HAIKU_MODEL, because it uses different slots for different tasks. Call other models with a Standard API Key at pay-as-you-go prices.

Error 1315

HTTP 429Fix, then resend

“This API Key is limited to enterprise coding package scenarios. Please go to the official website to replace the API Key of the corresponding product type.”

What it means

The key belongs to an enterprise coding package and was used outside it, for example for a general API call or from a tool the package does not cover.

How to fix it

Keep the enterprise key inside the package's coding tools and endpoints, and create a Standard API Key for everything else.

Error 1220

HTTP 403Fix, then resend

“You do not have permission to access ${API_name}”

What it means

The account or key is not allowed to use the API named in the message: a capability that has to be enabled or bought separately, or a key type that does not cover it.

How to fix it

Check what access the API needs on its page in Z.ai's docs, and the key's type on the API Keys page. If the account should have access, send Z.ai support the request id.

11 codes

GLM Coding Plan limits and status

The plan's allowances and its standing. Every limit message carries its own reset time, so read the message before changing anything.

Error 1308

HTTP 429Wait for the reset

“Usage limit reached for {number} {unit}. Your limit will reset at {next_flush_time}”

What it means

You have used a plan allowance for the period the message names. The current credit-based plans meter a rolling 5-hour window and a 7-day window, and the message says which one ran out and when it refills.

How to fix it

Wait for the reset time in the message. To make the allowance last: work outside Z.ai's weekday peak hours, when calls cost half the credits; give routine work to GLM-5.3-Flash; keep the context short. A move to a higher tier takes effect immediately.

Error 1310

HTTP 429Wait for the reset

“Weekly/Monthly Limit Exhausted. Your limit will reset at {next_flush_time}”

What it means

The longer allowance is spent. On the current plans that is the 7-day window; the reset time is in the message.

How to fix it

Wait for the reset time, move to a higher tier (immediate), or switch the tool to a Standard API Key and pay per token until the plan refills.

Error 1309

HTTP 429Fix, then resend

“Your GLM Coding Plan package has expired and is temporarily unavailable. You can resume using it after renewing the subscription on the official website.”

What it means

The subscription has ended, so the plan key has nothing to draw on. Z.ai's key notice says calls with an older Compatible key do not switch to pay-as-you-go by themselves when the plan expires.

How to fix it

Renew at z.ai/subscribe. To pay per token instead, create a Standard API Key, put it in the tool in place of the plan key (the endpoint can stay the same) and keep a balance on the account.

Error 1313

HTTP 429Fix, then resend

“Your account's current usage pattern does not comply with the Fair Usage Policy, and your request frequency has been limited. For details, please refer to the Subscription Service Agreement. To restore access, please submit a request.”

What it means

Z.ai's risk control has limited the account because its usage looks like something the plan's terms forbid: using the quota outside the supported tools, calling it from your own scripts or apps (the terms name SDK-based access), or sharing the account or key.

How to fix it

Stop the pattern, then file the restore request the message mentions. Keep the plan key inside supported coding tools and use a Standard API Key for anything you write yourself. Z.ai's usage policy says accounts with more than three violations may be banned.

Error 1314

HTTP 429Fix, then resend

“Your enterprise package has expired. Please contact your enterprise administrator.”

What it means

Your organisation's enterprise package has lapsed.

How to fix it

Ask your enterprise administrator to renew it.

Error 1316

HTTP 429Wait for the reset

“Usage limit reached for the past 5 hours. Insufficient balance for extra usage. Resets at {next_flush_time}.”

What it means

The 5-hour allowance is used up and the plan would continue on paid extra usage (Z.ai's on-demand overage, which its docs describe for Team plans), but the balance cannot pay for it.

How to fix it

Top up the balance that pays for overage, or wait for the reset time in the message.

Error 1317

HTTP 429Wait for the reset

“Usage limit reached for the past 7 days. Insufficient balance for extra usage. Resets at {next_flush_time}.”

What it means

The same as 1316 for the 7-day window: the weekly allowance is spent and the balance cannot pay for extra usage.

How to fix it

Top up the balance that pays for overage, or wait for the reset time in the message.

Error 1318

HTTP 429Wait for the reset

“Usage limit reached for the past 5 hours. Extra usage is not available due to monthly spend limit. Resets at {next_flush_time}.”

What it means

The 5-hour allowance is used up, and extra usage is switched on but has reached its monthly spending limit.

How to fix it

An administrator can raise the spending limit (Team plans set one per member); otherwise wait for the reset time in the message.

Error 1319

HTTP 429Wait for the reset

“Usage limit reached for the past 7 days. Extra usage is not available due to monthly spend limit. Resets at {next_flush_time}.”

What it means

The same as 1318 for the 7-day window.

How to fix it

An administrator can raise the spending limit; otherwise wait for the reset time in the message.

Error 1320

HTTP 429Wait for the reset

“Usage limit reached for the past 5 hours. Extra usage is not available due to monthly spend limit. Resets at {next_flush_time}.”

What it means

Z.ai's table gives 1320 exactly the same wording as 1318 and does not say how the two differ, so treat it the same way. With per-member limits on Team plans, check both the member's limit and any team-wide budget.

How to fix it

An administrator can raise the spending limit; otherwise wait for the reset time in the message.

Error 1321

HTTP 429Wait for the reset

“Usage limit reached for the past 7 days. Extra usage is not available due to monthly spend limit. Resets at {next_flush_time}.”

What it means

Worded exactly like 1319 in Z.ai's table, with no stated difference; treat it the same way.

How to fix it

An administrator can raise the spending limit; otherwise wait for the reset time in the message.

6 codes

Rate limits and capacity

The only errors worth retrying: back off, add jitter, and send fewer requests at once.

Error 1302

HTTP 429Retry with backoff

“Rate limit reached for requests”

What it means

Your account sent more requests, or more at the same time, than its limit allows. This one is about you, not Z.ai's capacity: the console's Rate Limits page (https://z.ai/manage-apikey/rate-limits) shows your account's limits.

How to fix it

Send fewer requests at once (in a coding agent, fewer parallel subagents or tool calls) and retry with exponential backoff and jitter instead of immediately.

Error 1305

HTTP 429Retry with backoff

“The service may be temporarily overloaded, please try again later”

What it means

Z.ai is short of capacity for that model right now. It is not about your account or your limits, and it is most common at busy hours and in the days after a launch.

How to fix it

Retry with exponential backoff. If it keeps happening, give routine work to GLM-5.3-Flash, or call the model through another host: OpenRouter routes GLM models across many providers, listed on our pricing page.

Error 1234

HTTP 500Retry with backoff

“Network error, error id: ${error_id}, please try again later”

What it means

A passing failure inside Z.ai's network.

How to fix it

Retry after a short wait. If it persists, send Z.ai support the error id from the message.

Error 1200

HTTP 500Retry with backoff

“API Call Error”

What it means

The call failed on Z.ai's side.

How to fix it

Retry with backoff; if the same request keeps failing, send Z.ai support the request id.

Error 1230

HTTP 500Retry with backoff

“API call process error”

What it means

Z.ai failed while processing the call.

How to fix it

Retry with backoff; if the same request keeps failing, send Z.ai support the request id.

HTTP 500

Retry with backoff

“Internal Error”

What it means

A plain HTTP 500 from Z.ai with no business code: an internal failure.

How to fix it

Retry with backoff.

Your account's limits (Z.ai console)Every host serving GLM, with prices

9 codes

Request errors

Something in the request does not match Z.ai's API. Most come from code written for OpenAI's API, whose ranges and model ids differ.

Error 1211

HTTP 400Fix, then resend

“Unknown Model, please check the model code.”

What it means

Z.ai does not recognise the model id. Its ids look like glm-5.3, glm-5.3-flash and glm-4.7-flash; OpenRouter's ids (z-ai/glm-5.3) are a different string and fail here, as does any id copied from another provider.

How to fix it

Use the exact id from Z.ai's docs. Our model pages list each model's Z.ai id and OpenRouter id side by side.

Error 1210

HTTP 400Fix, then resend

“Invalid API parameter, please check the documentation.”

What it means

Something in the request body does not fit Z.ai's API. The usual culprits are values OpenAI's API accepts and Z.ai's does not: Z.ai takes temperature only from 0 to 1 (OpenAI allows up to 2), top_p from 0.01 to 1, and max_tokens up to 131,072 on GLM-4.6 and later.

How to fix it

Compare the request with Z.ai's chat completion reference and remove or clamp the offending field.

Error 1214

HTTP 400Fix, then resend

“Parameter ${field} is invalid. Please check the documentation.”

What it means

The field named in the message has a value Z.ai does not allow. On GLM-5.3 and GLM-5.3-Flash watch the reasoning controls: thinking can only be enabled on them, and reasoning_effort supports only low, high and max.

How to fix it

Correct the field named in the message. For reasoning on GLM-5.3, send low, high or max, or leave reasoning_effort out (the default is max).

Error 1213

HTTP 400Fix, then resend

“Parameter ${field} was not received.”

What it means

A required field never arrived, often model or messages, or a field dropped by a proxy or SDK that rewrites the body.

How to fix it

Add the field named in the message, and check the body your code actually sends.

Error 1215

HTTP 400Fix, then resend

“${field1} and ${field2} cannot be set simultaneously, please check the documentation”

What it means

Two fields that cannot be combined were both sent; the message names them.

How to fix it

Send only one of the two.

Error 1212

HTTP 400Fix, then resend

“Current model does not support ${method} call method”

What it means

The model exists but not for the way it was called, for example an image or video input sent to a text-only model, or an asynchronous call to a model that only answers synchronously.

How to fix it

Use a model that supports the method. GLM-5.3-Flash accepts text, image and video input.

Error 1222

HTTP 400Fix, then resend

“API ${API_name} does not exist”

What it means

The path matches no Z.ai API, usually because the base URL is wrong. Z.ai's bases are https://api.z.ai/api/paas/v4 (pay-as-you-go, OpenAI format), https://api.z.ai/api/coding/paas/v4 (Coding Plan, OpenAI format) and https://api.z.ai/api/anthropic (Anthropic format).

How to fix it

Set the base URL exactly as written. OpenAI SDKs add /chat/completions to it themselves, so do not include it.

Error 1221

HTTP 400Fix, then resend

“API ${API_name} has been taken offline”

What it means

The API or model you called has been retired.

How to fix it

Move to its replacement. Z.ai's migration guide covers moving to GLM-5.3 (docs.z.ai/guides/overview/migrate-to-glm-new).

Error 1261

HTTP 400Fix, then resend

“Prompt too long”

What it means

The input is longer than the model's context window: 1M on GLM-5.3, GLM-5.3-Flash and GLM-5.2, 200K on GLM-5.1, GLM-5, GLM-4.7-Flash and GLM-4.7.

How to fix it

Trim or summarise the history (in Claude Code, /compact), split the input, or use a 1M-context model. In a streamed response the same condition ends with finish_reason model_context_window_exceeded.

Z.ai chat completion reference Every GLM model, with its Z.ai and OpenRouter ids

4 codes

Authentication

The key never got as far as billing. All of these are HTTP 401.

Error 1000

HTTP 401Fix, then resend

“Authentication Failed”

What it means

Z.ai received a key and did not accept it: a typo, a key that was deleted or regenerated, a stray space or line break pasted with it, or a key from another platform.

How to fix it

Copy the key again from Z.ai's API Keys page and send it as Authorization: Bearer <key>. In Claude Code the key goes in ANTHROPIC_AUTH_TOKEN, with ANTHROPIC_BASE_URL set to https://api.z.ai/api/anthropic.

Error 1001

HTTP 401Fix, then resend

“Authentication parameter not received in Header, unable to authenticate”

What it means

The request arrived with no key at all, usually because the environment variable your tool reads is empty, misspelled, or set in a different shell or profile than the one the tool runs in.

How to fix it

Check the variable in the same shell the tool runs in, or put the key in the tool's own settings file, then restart the tool so it reads it.

Error 1003

HTTP 401Fix, then resend

“Authentication Token expired, please regenerate/obtain”

What it means

The credential was a token with an expiry time and that time has passed, typically in code that signs its own short-lived token from the key instead of sending the key.

How to fix it

Generate a fresh token, or send the API key itself as the Bearer value, as Z.ai's quick-start does.

Error 1005

HTTP 401Fix, then resend

“Need Two-Factor Authentication”

What it means

Z.ai wants the account to pass two-factor authentication before it serves the key.

How to fix it

Sign in at z.ai and complete the two-factor step it asks for. The error is about the account, so a new key will not get around it.

1 codes

Content filter

Z.ai's own content policy, applied to the input and to the output.

Error 1301

HTTP 400Fix, then resend

“System detected potentially unsafe or sensitive content in input or generation. Please avoid using prompts that may generate sensitive content. Thank you for your cooperation.”

What it means

Z.ai's content filter flagged the input or the output. In a streamed response the same thing ends the stream with finish_reason sensitive instead of this code.

How to fix it

Rephrase or remove the passage that triggered it. False positives happen on ordinary technical material; if one blocks legitimate work, report it to Z.ai support with the request id.

Streaming

When there is no error code

In a streamed (SSE) call, a failure during inference does not return any of the codes above. The stream ends and the reason is in finish_reason, so check it on every streamed response, not only the HTTP status.

Z.ai chat completion reference

stop: The model finished normally.
tool_calls: The model stopped to call a tool.
length: The output hit max_tokens. Raise it (up to 131,072) or ask for less.
model_context_window_exceeded: The conversation outgrew the context window; the streamed form of 1261.
sensitive: The content filter cut the response; the streamed form of 1301.
network_error: Inference broke off inside Z.ai; retry, as with 1234.

FAQ

Quick answers on Z.ai errors

What does Z.ai error 1113 mean?

"Insufficient balance or no resource package" means Z.ai billed the call to your pay-as-you-go balance and the balance is empty. If you have a GLM Coding Plan, the call did not qualify for the plan: the usual cause since October 16, 2026 is a Standard API Key, which always bills the balance. Use a key from API Keys, Personal Coding Plan, at a coding endpoint (https://api.z.ai/api/anthropic for Claude Code). Without a plan, top up or use a free model (GLM-4.7-Flash and GLM-4.5-Flash).

Why do I get error 1113 after buying the GLM Coding Plan?

Because the request is not one the plan pays for. Three conditions must all hold: a Coding Plan key (not a Standard API Key), a coding endpoint (https://api.z.ai/api/anthropic or https://api.z.ai/api/coding/paas/v4, not https://api.z.ai/api/paas/v4), and a plan model (GLM-5.3 or GLM-5.3-Flash) called from a supported coding tool. Miss one and the call is billed to your balance, which returns 1113 when it is empty.

What is the difference between Z.ai errors 1302 and 1305?

1302 is your account's own rate limit: you sent too many requests or too many at once, and your limits are on the console's Rate Limits page (https://z.ai/manage-apikey/rate-limits). 1305 is Z.ai's capacity: the model is overloaded for everyone. Both are HTTP 429 and both are worth retrying with exponential backoff; for 1302 also send fewer requests in parallel.

When does a GLM Coding Plan limit reset after error 1308 or 1310?

The reset time is in the error message itself ({next_flush_time}). The current credit-based plans meter a rolling 5-hour window and a 7-day window: 1308 names the period that ran out, 1310 is the longer one. Calls outside Z.ai's weekday peak hours cost half the credits, so the allowance lasts longer off-peak.

Why does Claude Code say my Coding Plan does not include the model (error 1311)?

The plan covers only GLM-5.3 and GLM-5.3-Flash (requests for GLM-5.2 and GLM-5.1 are served by GLM-5.3, and GLM-4.7 by GLM-5.3-Flash). Set ANTHROPIC_DEFAULT_OPUS_MODEL, ANTHROPIC_DEFAULT_SONNET_MODEL and ANTHROPIC_DEFAULT_HAIKU_MODEL to glm-5.3 or glm-5.3-flash, since Claude Code uses different slots for different tasks.

Which Z.ai API errors are safe to retry?

Retry 1302 and 1305 (both HTTP 429), 1200, 1230 and 1234, and a plain HTTP 500, with exponential backoff and jitter. Wait for the reset time in the message for the plan limits 1308, 1310 and 1316 to 1321. Everything else, including 1113 and 1311 although they are HTTP 429, fails again until you change the key, the model, the endpoint or the request.

Why don't Z.ai error codes appear in streamed responses?

Once a stream has started, Z.ai cannot change the HTTP status, so a failure during inference ends the stream with a finish_reason instead: model_context_window_exceeded (the streamed form of 1261), sensitive (1301), network_error (retry it) or length (the output hit max_tokens).

GLM Coding Plan: prices, limits and setup Calling GLM from code: endpoints and free options Z.ai, the lab behind GLM

Build something with GLM

Past the errors, the next step is shipping a real product on a model like GLM, without wiring up the infrastructure yourself.

Build something on Founden

Ox Alpha

Z.ai API errors

Z.ai API error codes, and how to fix each one

Independent reference, not affiliated with Z.aiUpdated for the October 16 API key split

The six you are most likely seeing

1113Fix, then resend

Call billed to an empty balance

Use a Coding Plan key at a coding endpoint, or top up the balance.

Read the fix

1302Retry with backoff

Your account's rate limit

Send fewer requests at once and retry with backoff.

Read the fix

1305Retry with backoff

Z.ai is overloaded

Retry with exponential backoff, or call the model through another host.

Read the fix

1308Wait for the reset

Plan allowance used up

Wait for the reset time in the message; off-peak calls cost half the credits.

Read the fix

1311Fix, then resend

Model not in your Coding Plan

Set every model in your tool to glm-5.3 or glm-5.3-flash.

Read the fix

1261Fix, then resend

Input longer than the context window

Trim the history (/compact in Claude Code) or use a 1M-context model.

Read the fix

GLM Coding Plan

Four things a plan call needs

Z.ai API key notice, Oct 16, 2026 Z.ai API Keys page

The full GLM Coding Plan guide, with setup

The key: A key from API Keys, Personal Coding Plan (Team members: Team Coding Plan, My Plan). Since October 16, 2026 a Standard API Key never draws on the plan.
The endpoint: https://api.z.ai/api/anthropic for Claude Code, https://api.z.ai/api/coding/paas/v4 for OpenAI-compatible tools. Not https://api.z.ai/api/paas/v4, which bills the balance.
The model: glm-5.3 or glm-5.3-flash, the only two the plan calls (older ids are rerouted to them).
The tool: One of Z.ai's supported coding tools. Your own scripts and apps need a Standard API Key.

Reading an error

Two codes in every error

HTTP/2 429
{"error":{"code":"1113","message":"Insufficient balance or no resource package. Please recharge."}}

Coding tools show the same JSON inside their own error line, so look for the four-digit number after code.
Z.ai's example sends the code as a string ("1113") though its schema calls it an integer, so compare it as a string in your own error handling.
OpenAI's and Anthropic's SDKs retry a 429 twice by default, and coding agents retry too, so a 1113 or a plan limit can look like a hang before it surfaces.
Once a response is streaming, failures arrive as a finish_reason instead (see below).

What to do

Retry, wait, or fix

Retry with exponential backoff and jitter

Rate limits, overload and Z.ai's own server errors. For 1302, also send fewer requests at once.

1302 1305 1234 1200 1230 HTTP 500

Wait for the reset time in the message

Coding Plan limits. The message states when the allowance refills; retrying before then fails again.

1308 1310 1316 1317 1318 1319 1320 1321

Fix something, then resend

The key, the endpoint, the model, the plan or the request is wrong. These fail every time until it changes, including 1113 and 1311 although they arrive as HTTP 429.

1113 1311 1315 1220 1309 1313 1314 1211 1210 1214 1213 1215 1212 1222 1221 1261 1000 1001 1003 1005 1301

Every code

The full lookup table

Z.ai's complete list, in code order. Each code links to its explanation below.

Code	HTTP	What it means	What to do
1000	401	API key rejected	Fix, then resend
1001	401	No key in the request	Fix, then resend
1003	401	Token expired	Fix, then resend
1005	401	Account needs two-factor	Fix, then resend
1113	429	Call billed to an empty balance	Fix, then resend
1200	500	Server-side call error	Retry with backoff
1210	400	Invalid parameter	Fix, then resend
1211	400	Model id not recognised	Fix, then resend
1212	400	Model does not support this call	Fix, then resend
1213	400	Required field missing	Fix, then resend
1214	400	A field's value is not allowed	Fix, then resend
1215	400	Two fields that conflict	Fix, then resend
1220	403	No access to this API	Fix, then resend
1221	400	API or model retired	Fix, then resend
1222	400	Wrong path or base URL	Fix, then resend
1230	500	Server-side processing error	Retry with backoff
1234	500	Transient network error	Retry with backoff
1261	400	Input longer than the context window	Fix, then resend
1301	400	Content filter triggered	Fix, then resend
1302	429	Your account's rate limit	Retry with backoff
1305	429	Z.ai is overloaded	Retry with backoff
1308	429	Plan allowance used up	Wait for the reset
1309	429	Coding Plan expired	Fix, then resend
1310	429	Weekly or monthly limit used up	Wait for the reset
1311	429	Model not in your Coding Plan	Fix, then resend
1313	429	Throttled under the fair-use rules	Fix, then resend
1314	429	Enterprise package expired	Fix, then resend
1315	429	Enterprise coding key used elsewhere	Fix, then resend
1316	429	5-hour limit, no balance for overage	Wait for the reset
1317	429	7-day limit, no balance for overage	Wait for the reset
1318	429	5-hour limit, overage capped	Wait for the reset
1319	429	7-day limit, overage capped	Wait for the reset
1320	429	5-hour limit, overage capped	Wait for the reset
1321	429	7-day limit, overage capped	Wait for the reset
HTTP 500	500	Internal error (no business code)	Retry with backoff

Z.ai error table (docs.z.ai)

4 codes

Balance, key type and plan access

The errors behind most 'it worked yesterday' reports since Z.ai split its API keys on October 16, 2026. They arrive as HTTP 429 or 403, but none of them clears by retrying.

Error 1113

HTTP 429Fix, then resend

“Insufficient balance or no resource package. Please recharge.”

What it means

How to fix it

Error 1311

HTTP 429Fix, then resend

“Your current subscription plan does not yet include access to ${model_name}”

What it means

How to fix it

Error 1315

HTTP 429Fix, then resend

“This API Key is limited to enterprise coding package scenarios. Please go to the official website to replace the API Key of the corresponding product type.”

What it means

The key belongs to an enterprise coding package and was used outside it, for example for a general API call or from a tool the package does not cover.

How to fix it

Keep the enterprise key inside the package's coding tools and endpoints, and create a Standard API Key for everything else.

Error 1220

HTTP 403Fix, then resend

“You do not have permission to access ${API_name}”

What it means

The account or key is not allowed to use the API named in the message: a capability that has to be enabled or bought separately, or a key type that does not cover it.

How to fix it

Check what access the API needs on its page in Z.ai's docs, and the key's type on the API Keys page. If the account should have access, send Z.ai support the request id.

11 codes

GLM Coding Plan limits and status

The plan's allowances and its standing. Every limit message carries its own reset time, so read the message before changing anything.

Error 1308

HTTP 429Wait for the reset

“Usage limit reached for {number} {unit}. Your limit will reset at {next_flush_time}”

What it means

How to fix it

Error 1310

HTTP 429Wait for the reset

“Weekly/Monthly Limit Exhausted. Your limit will reset at {next_flush_time}”

What it means

The longer allowance is spent. On the current plans that is the 7-day window; the reset time is in the message.

How to fix it

Wait for the reset time, move to a higher tier (immediate), or switch the tool to a Standard API Key and pay per token until the plan refills.

Error 1309

HTTP 429Fix, then resend

“Your GLM Coding Plan package has expired and is temporarily unavailable. You can resume using it after renewing the subscription on the official website.”

What it means

The subscription has ended, so the plan key has nothing to draw on. Z.ai's key notice says calls with an older Compatible key do not switch to pay-as-you-go by themselves when the plan expires.

How to fix it

Renew at z.ai/subscribe. To pay per token instead, create a Standard API Key, put it in the tool in place of the plan key (the endpoint can stay the same) and keep a balance on the account.

Error 1313

HTTP 429Fix, then resend

What it means

How to fix it

Error 1314

HTTP 429Fix, then resend

“Your enterprise package has expired. Please contact your enterprise administrator.”

What it means

Your organisation's enterprise package has lapsed.

How to fix it

Ask your enterprise administrator to renew it.

Error 1316

HTTP 429Wait for the reset

“Usage limit reached for the past 5 hours. Insufficient balance for extra usage. Resets at {next_flush_time}.”

What it means

The 5-hour allowance is used up and the plan would continue on paid extra usage (Z.ai's on-demand overage, which its docs describe for Team plans), but the balance cannot pay for it.

How to fix it

Top up the balance that pays for overage, or wait for the reset time in the message.

Error 1317

HTTP 429Wait for the reset

“Usage limit reached for the past 7 days. Insufficient balance for extra usage. Resets at {next_flush_time}.”

What it means

The same as 1316 for the 7-day window: the weekly allowance is spent and the balance cannot pay for extra usage.

How to fix it

Top up the balance that pays for overage, or wait for the reset time in the message.

Error 1318

HTTP 429Wait for the reset

“Usage limit reached for the past 5 hours. Extra usage is not available due to monthly spend limit. Resets at {next_flush_time}.”

What it means

The 5-hour allowance is used up, and extra usage is switched on but has reached its monthly spending limit.

How to fix it

An administrator can raise the spending limit (Team plans set one per member); otherwise wait for the reset time in the message.

Error 1319

HTTP 429Wait for the reset

“Usage limit reached for the past 7 days. Extra usage is not available due to monthly spend limit. Resets at {next_flush_time}.”

What it means

The same as 1318 for the 7-day window.

How to fix it

An administrator can raise the spending limit; otherwise wait for the reset time in the message.

Error 1320

HTTP 429Wait for the reset

“Usage limit reached for the past 5 hours. Extra usage is not available due to monthly spend limit. Resets at {next_flush_time}.”

What it means

How to fix it

An administrator can raise the spending limit; otherwise wait for the reset time in the message.

Error 1321

HTTP 429Wait for the reset

“Usage limit reached for the past 7 days. Extra usage is not available due to monthly spend limit. Resets at {next_flush_time}.”

What it means

Worded exactly like 1319 in Z.ai's table, with no stated difference; treat it the same way.

How to fix it

An administrator can raise the spending limit; otherwise wait for the reset time in the message.

6 codes

Rate limits and capacity

The only errors worth retrying: back off, add jitter, and send fewer requests at once.

Error 1302

HTTP 429Retry with backoff

“Rate limit reached for requests”

What it means

How to fix it

Send fewer requests at once (in a coding agent, fewer parallel subagents or tool calls) and retry with exponential backoff and jitter instead of immediately.

Error 1305

HTTP 429Retry with backoff

“The service may be temporarily overloaded, please try again later”

What it means

Z.ai is short of capacity for that model right now. It is not about your account or your limits, and it is most common at busy hours and in the days after a launch.

How to fix it

Error 1234

HTTP 500Retry with backoff

“Network error, error id: ${error_id}, please try again later”

What it means

A passing failure inside Z.ai's network.

How to fix it

Retry after a short wait. If it persists, send Z.ai support the error id from the message.

Error 1200

HTTP 500Retry with backoff

“API Call Error”

What it means

The call failed on Z.ai's side.

How to fix it

Retry with backoff; if the same request keeps failing, send Z.ai support the request id.

Error 1230

HTTP 500Retry with backoff

“API call process error”

What it means

Z.ai failed while processing the call.

How to fix it

Retry with backoff; if the same request keeps failing, send Z.ai support the request id.

HTTP 500

Retry with backoff

“Internal Error”

What it means

A plain HTTP 500 from Z.ai with no business code: an internal failure.

How to fix it

Retry with backoff.

Your account's limits (Z.ai console)Every host serving GLM, with prices

9 codes

Request errors

Something in the request does not match Z.ai's API. Most come from code written for OpenAI's API, whose ranges and model ids differ.

Error 1211

HTTP 400Fix, then resend

“Unknown Model, please check the model code.”

What it means

How to fix it

Use the exact id from Z.ai's docs. Our model pages list each model's Z.ai id and OpenRouter id side by side.

Error 1210

HTTP 400Fix, then resend

“Invalid API parameter, please check the documentation.”

What it means

How to fix it

Compare the request with Z.ai's chat completion reference and remove or clamp the offending field.

Error 1214

HTTP 400Fix, then resend

“Parameter ${field} is invalid. Please check the documentation.”

What it means

How to fix it

Correct the field named in the message. For reasoning on GLM-5.3, send low, high or max, or leave reasoning_effort out (the default is max).

Error 1213

HTTP 400Fix, then resend

“Parameter ${field} was not received.”

What it means

A required field never arrived, often model or messages, or a field dropped by a proxy or SDK that rewrites the body.

How to fix it

Add the field named in the message, and check the body your code actually sends.

Error 1215

HTTP 400Fix, then resend

“${field1} and ${field2} cannot be set simultaneously, please check the documentation”

What it means

Two fields that cannot be combined were both sent; the message names them.

How to fix it

Send only one of the two.

Error 1212

HTTP 400Fix, then resend

“Current model does not support ${method} call method”

What it means

The model exists but not for the way it was called, for example an image or video input sent to a text-only model, or an asynchronous call to a model that only answers synchronously.

How to fix it

Use a model that supports the method. GLM-5.3-Flash accepts text, image and video input.

Error 1222

HTTP 400Fix, then resend

“API ${API_name} does not exist”

What it means

How to fix it

Set the base URL exactly as written. OpenAI SDKs add /chat/completions to it themselves, so do not include it.

Error 1221

HTTP 400Fix, then resend

“API ${API_name} has been taken offline”

What it means

The API or model you called has been retired.

How to fix it

Move to its replacement. Z.ai's migration guide covers moving to GLM-5.3 (docs.z.ai/guides/overview/migrate-to-glm-new).

Error 1261

HTTP 400Fix, then resend

“Prompt too long”

What it means

The input is longer than the model's context window: 1M on GLM-5.3, GLM-5.3-Flash and GLM-5.2, 200K on GLM-5.1, GLM-5, GLM-4.7-Flash and GLM-4.7.

How to fix it

Z.ai chat completion reference Every GLM model, with its Z.ai and OpenRouter ids

4 codes

Authentication

The key never got as far as billing. All of these are HTTP 401.

Error 1000

HTTP 401Fix, then resend

“Authentication Failed”

What it means

Z.ai received a key and did not accept it: a typo, a key that was deleted or regenerated, a stray space or line break pasted with it, or a key from another platform.

How to fix it

Error 1001

HTTP 401Fix, then resend

“Authentication parameter not received in Header, unable to authenticate”

What it means

The request arrived with no key at all, usually because the environment variable your tool reads is empty, misspelled, or set in a different shell or profile than the one the tool runs in.

How to fix it

Check the variable in the same shell the tool runs in, or put the key in the tool's own settings file, then restart the tool so it reads it.

Error 1003

HTTP 401Fix, then resend

“Authentication Token expired, please regenerate/obtain”

What it means

The credential was a token with an expiry time and that time has passed, typically in code that signs its own short-lived token from the key instead of sending the key.

How to fix it

Generate a fresh token, or send the API key itself as the Bearer value, as Z.ai's quick-start does.

Error 1005

HTTP 401Fix, then resend

“Need Two-Factor Authentication”

What it means

Z.ai wants the account to pass two-factor authentication before it serves the key.

How to fix it

Sign in at z.ai and complete the two-factor step it asks for. The error is about the account, so a new key will not get around it.

1 codes

Content filter

Z.ai's own content policy, applied to the input and to the output.

Error 1301

HTTP 400Fix, then resend

“System detected potentially unsafe or sensitive content in input or generation. Please avoid using prompts that may generate sensitive content. Thank you for your cooperation.”

What it means

Z.ai's content filter flagged the input or the output. In a streamed response the same thing ends the stream with finish_reason sensitive instead of this code.

How to fix it

Rephrase or remove the passage that triggered it. False positives happen on ordinary technical material; if one blocks legitimate work, report it to Z.ai support with the request id.

Streaming

When there is no error code

Z.ai chat completion reference

stop: The model finished normally.
tool_calls: The model stopped to call a tool.
length: The output hit max_tokens. Raise it (up to 131,072) or ask for less.
model_context_window_exceeded: The conversation outgrew the context window; the streamed form of 1261.
sensitive: The content filter cut the response; the streamed form of 1301.
network_error: Inference broke off inside Z.ai; retry, as with 1234.

FAQ

Quick answers on Z.ai errors

What does Z.ai error 1113 mean?

Why do I get error 1113 after buying the GLM Coding Plan?

What is the difference between Z.ai errors 1302 and 1305?

When does a GLM Coding Plan limit reset after error 1308 or 1310?

Why does Claude Code say my Coding Plan does not include the model (error 1311)?

Which Z.ai API errors are safe to retry?

Why don't Z.ai error codes appear in streamed responses?

GLM Coding Plan: prices, limits and setup Calling GLM from code: endpoints and free options Z.ai, the lab behind GLM

Build something with GLM

Past the errors, the next step is shipping a real product on a model like GLM, without wiring up the infrastructure yourself.

Build something on Founden

Z.ai API Error Codes Explained (1113, 1302, 1305) | Ox Alpha