What usually causes a 404 Not Found?

Most often the Base URL does not match the protocol: an OpenAI-protocol URL missing /v1, an Anthropic URL that wrongly includes /v1, or a Gemini path that does not use the full /v1beta/models/... endpoint. Re-check the Base URL and endpoint against the API Protocols page.

Should I retry on a 5xx, or contact support directly?

A 5xx is a transient upstream or relay failure. Retry a limited number of times with exponential backoff and save the request ID from the response headers. If the same kind of request keeps failing within 5 minutes, send the request ID to support — never send your full API key.

Error Codes

LMU AI error-code reference: meanings and fixes for 401 / 403 / 404 / 429 / 5xx statuses, batch-image business codes, and lookup from raw error text to a fix.

This page collects the error codes scattered across the individual API docs into one cheat sheet. First locate the broad class by HTTP status code, then find the specific fix by the raw error text.

Record the request ID before troubleshooting

The request ID in the response headers is the single most useful piece of information for diagnosing a problem. When you contact support, include the request ID and the full error text — do not send your full API key.

HTTP status codes

Status	Typical cause	What to do
`400`	Invalid request body, parameter, model ID or model path; bad image format / encoding	Fix the request — do not just retry; switching accounts or retrying will not succeed
`401`	API key missing, invalid, or disabled; Base URL does not match the protocol; the IDE was not restarted so the old config is still in effect	Check the key and Base URL (see API Protocols), restart the IDE; details in Issue 3
`402`	Insufficient balance	Top up or reduce the workload
`403`	Two possibilities, both possible: ① the API key's group does not have image generation enabled; ② insufficient balance, subscription, billing eligibility, or permission	First confirm the account balance and subscription status, then confirm the group has the capability; if both are fine and it still errors, contact support
`404`	OpenAI-protocol URL missing `/v1`, Anthropic URL wrongly including `/v1`, Gemini not using `/v1beta/models/...`; or the endpoint is not supported by the current group	Re-check the Base URL and full endpoint against API Protocols
`413`	Image-to-image request body too large	Compress the input image
`429`	① Daily quota exhausted; ② concurrency, RPM, or upstream quota limited	For an exhausted quota see Issue 2; for rate limiting, back off exponentially and lower concurrency and RPM
`500`	Internal or capacity error	Record the error code and request ID, retry a limited number of times
`502`	Upstream auth, permission, or service temporarily failed	Back off and retry; contact support if needed
`503`	No available upstream account or upstream overloaded; may also be an environment variable overriding the key	First check env vars (see Issue 6), then retry with delay and lower traffic
`504`	Gateway or upstream timeout	Re-issue as a fresh, independent request

What to retry, and what must be fixed instead

Do not retry: 400, 401, and any explicit balance, permission, parameter, or content-policy error — the request itself is invalid, and any upstream account returns the same result. You must fix the request first.
Safe to retry: 429, 502, 503, 504, and network drops, connection resets, and read timeouts — all transient failures. Use exponential backoff for a limited number of retries (2–3 recommended) to avoid paying for repeated calls; on 429 also lower concurrency and RPM.

This classification comes from the retry guidance in the image APIs — see Gemini Image · Retry guidance.

Full error descriptions for each API: Gemini Image, GPT Image, Grok Image, Gemini Batch Image.

Business error codes (batch image)

Beyond HTTP status codes, the Gemini Batch Image API also returns business error codes, shown in the console as error code: BATCH_IMAGE_XXX plus a request ID.

Error code	HTTP	Meaning	What to do
`BATCH_IMAGE_DISABLED`	404	Batch image generation disabled globally	Send the error code and request ID to the administrator
`BATCH_IMAGE_GROUP_DISABLED`	403	The current key's group does not allow batch image or is not a Gemini group	Change the key or ask the administrator to enable group permission
`BATCH_IMAGE_NO_ACCOUNT_AVAILABLE`	502	No batch-execution resource currently available	Save the request ID and contact the administrator
`BATCH_IMAGE_SETTLEMENT_PRICING_MISSING`	400	The batch model has no billing price	Ask the administrator to configure the model price
`BATCH_IMAGE_INVALID_MODEL`	400	No model provided	Use a model from the batch model list
`BATCH_IMAGE_INVALID_ITEMS`	400	Invalid items, resolution, or request fields	Check the request body; only 1K is supported for now
`BATCH_IMAGE_DUPLICATE_CUSTOM_ID`	400	Duplicate `custom_id`	Ensure it is unique within the batch
`BATCH_IMAGE_PROMPT_TOO_LONG`	400	Prompt too long	Shorten the prompt
`BATCH_IMAGE_TOO_MANY_OUTPUT_IMAGES`	400	Image count after expansion exceeds the limit	Reduce items or output_count
`BATCH_IMAGE_INVALID_REFERENCE_IMAGE`	400	Invalid reference-image format, size, or URI	Check MIME, Base64, and file_uri
`BATCH_IMAGE_INSUFFICIENT_BALANCE`	402	Balance too low to reserve the charge	Top up or reduce the workload
`BATCH_IMAGE_IDEMPOTENCY_CONFLICT`	409	The same idempotency key maps to a different request body	Use a new Idempotency-Key
`BATCH_IMAGE_PROVIDER_SUBMIT_FAILED`	502	Upstream batch-task creation failed	Save the request ID, retry a limited number of times, or contact the administrator
`BATCH_IMAGE_QUEUE_FAILED`	502	The async task service is temporarily unavailable	Save the request ID and contact the administrator
`BATCH_IMAGE_NOT_READY`	409	Tried to download before the task completed	Wait until the status becomes completed
`BATCH_IMAGE_OUTPUT_DELETED`	410	The output was already cleaned up	It can no longer be downloaded; create the task again
`BATCH_IMAGE_ITEM_FAILED`	409	The specified task item has no successful image	Check item.error
`BATCH_IMAGE_DOWNLOAD_LIMITED`	429	Too many simultaneous downloads	Retry later

Look up by raw error text

Match the error text you actually see against the table below and jump straight to the detailed fix.

Raw error text	Meaning	Detailed fix
`stream disconnected before completion`	Stream dropped, usually a VPN / proxy / system proxy rotating the exit IP	Issue 1
`exceeded retry limit, last status: 429 Too Many Requests`	Daily quota exhausted	Issue 2
`401 Unauthorized: Incorrect API key provided`	The request still went to the official endpoint, not the LMU AI relay	Issue 3
`running scripts is disabled on this system` (Windows)	PowerShell execution policy restriction	Issue 4
`CODEX is not recognized as a cmdlet` (Windows)	Node.js not installed or PATH broken	Issue 5
`503 No available accounts`	A shell env var overrides the key configured in the IDE	Issue 6
`400 Invalid signature in thinking block`	Switching models across groups in one conversation; the thinking signature cannot be verified	Issue 7
`400 Unknown parameter: 'tools[0].n'`	A `tools` parameter was wrongly sent to an image endpoint	Issue 8
`No available accounts` / model unavailable	Called a model outside the current group's available range	API Protocols → Troubleshooting

Errors on the Usage export API

The Usage export API uses JWT authentication, so its error semantics differ from the API-key channel above:

Status	Cause	What to do
`401`	JWT expired	Renew with the `refresh_token`, or sign in again
`403`	Unauthorized access (e.g. querying an `api_key_id` that is not yours)	Check whether the key belongs to the current account
`400`	Bad parameter (e.g. wrong `start_date` format)	Check the `YYYY-MM-DD` format and that `timezone` is valid

Still stuck?

Step-by-step worked cases are in the FAQ
Base URL / endpoint rules are in API Protocols
Protection against a stolen key is in Key Security
When contacting support, include the request ID and the full error text

Last updated: 2026-08-26