Using Batch API to generate embeddings but getting 429 RESOURCE_EXHAUSTED errors in Paid Tier 1

Sushant_Karki · September 19, 2025, 4:07pm

Hi,
I am using the Batch api to generate embeddings but getting 429 RESOURCE_EXHAUSTED on even just 1 request. I am using similar code to the Batch API cookbook that is mentioned in the gemini embedding documentation.

My file is not too big:
Size: 3.4 MB
Number of lines (jsonlines): 1937
Maximum token count: 706226 (I used the count_tokens API on the entire jsonl file, not just the text to be embedded)

Here are some logs:

INFO:httpx:HTTP Request: POST https://generativelanguage.googleapis.com/upload/v1beta/files "HTTP/1.1 200 OK"
INFO:httpx:HTTP Request: POST https://generativelanguage.googleapis.com/upload/v1beta/files?upload_id=xxxxxxxxx&upload_protocol=resumable "HTTP/1.1 200 OK"
/tmp/ipykernel_133/2660981119.py:56: ExperimentalWarning: batches.create_embeddings() is experimental and may change without notice.
INFO:httpx:HTTP Request: POST https://generativelanguage.googleapis.com/v1beta/models/gemini-embedding-001:asyncBatchEmbedContent "HTTP/1.1 429 Too Many Requests"
---------------------------------------------------------------------------
ClientError                               Traceback (most recent call last)
Cell In[99], line 56
     48 uploaded_file = client.files.upload(
     49     file=buffer,
     50     config=types.UploadFileConfig(
   (...)
     53     )
     54 )
     55 print(uploaded_file.name)
---> 56 batch_job = client.batches.create_embeddings(
     57     model="gemini-embedding-001",
     58     src=types.EmbeddingsBatchJobSource(file_name=uploaded_file.name)
     59 )
     60 print(batch_job)
     61 batch_jobs.append(batch_job)

File /opt/conda/lib/python3.11/site-packages/google/genai/batches.py:2667, in Batches.create_embeddings(self, model, src, config)
   2665   raise ValueError('Vertex AI does not support batches.create_embeddings.')
   2666 elif src.inlined_requests is None:
-> 2667   return self._create_embeddings(model=model, src=src, config=config)
   2669 path, request_dict = _create_inlined_embedding_request_dict(
   2670     self._api_client, parameter_model
   2671 )
   2673 response = self._api_client.request(
   2674     'post', path, request_dict, http_options
   2675 )

File /opt/conda/lib/python3.11/site-packages/google/genai/batches.py:2235, in Batches._create_embeddings(self, model, src, config)
   2232 request_dict = _common.convert_to_dict(request_dict)
   2233 request_dict = _common.encode_unserializable_types(request_dict)
-> 2235 response = self._api_client.request(
   2236     'post', path, request_dict, http_options
   2237 )
   2239 response_dict = '' if not response.body else json.loads(response.body)
   2241 if not self._api_client.vertexai:

File /opt/conda/lib/python3.11/site-packages/google/genai/_api_client.py:1290, in BaseApiClient.request(self, http_method, path, request_dict, http_options)
   1280 def request(
   1281     self,
   1282     http_method: str,
   (...)
   1285     http_options: Optional[HttpOptionsOrDict] = None,
   1286 ) -> SdkHttpResponse:
   1287   http_request = self._build_request(
   1288       http_method, path, request_dict, http_options
   1289   )
-> 1290   response = self._request(http_request, http_options, stream=False)
   1291   response_body = (
   1292       response.response_stream[0] if response.response_stream else ''
   1293   )
   1294   return SdkHttpResponse(headers=response.headers, body=response_body)

File /opt/conda/lib/python3.11/site-packages/google/genai/_api_client.py:1126, in BaseApiClient._request(self, http_request, http_options, stream)
   1123     retry = tenacity.Retrying(**retry_kwargs)
   1124     return retry(self._request_once, http_request, stream)  # type: ignore[no-any-return]
-> 1126 return self._retry(self._request_once, http_request, stream)

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:477, in Retrying.__call__(self, fn, *args, **kwargs)
    475 retry_state = RetryCallState(retry_object=self, fn=fn, args=args, kwargs=kwargs)
    476 while True:
--> 477     do = self.iter(retry_state=retry_state)
    478     if isinstance(do, DoAttempt):
    479         try:

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:378, in BaseRetrying.iter(self, retry_state)
    376 result = None
    377 for action in self.iter_state.actions:
--> 378     result = action(retry_state)
    379 return result

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:420, in BaseRetrying._post_stop_check_actions.<locals>.exc_check(rs)
    418 retry_exc = self.retry_error_cls(fut)
    419 if self.reraise:
--> 420     raise retry_exc.reraise()
    421 raise retry_exc from fut.exception()

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:187, in RetryError.reraise(self)
    185 def reraise(self) -> t.NoReturn:
    186     if self.last_attempt.failed:
--> 187         raise self.last_attempt.result()
    188     raise self

File /opt/conda/lib/python3.11/concurrent/futures/_base.py:449, in Future.result(self, timeout)
    447     raise CancelledError()
    448 elif self._state == FINISHED:
--> 449     return self.__get_result()
    451 self._condition.wait(timeout)
    453 if self._state in [CANCELLED, CANCELLED_AND_NOTIFIED]:

File /opt/conda/lib/python3.11/concurrent/futures/_base.py:401, in Future.__get_result(self)
    399 if self._exception:
    400     try:
--> 401         raise self._exception
    402     finally:
    403         # Break a reference cycle with the exception in self._exception
    404         self = None

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:480, in Retrying.__call__(self, fn, *args, **kwargs)
    478 if isinstance(do, DoAttempt):
    479     try:
--> 480         result = fn(*args, **kwargs)
    481     except BaseException:  # noqa: B902
    482         retry_state.set_exception(sys.exc_info())  # type: ignore[arg-type]

File /opt/conda/lib/python3.11/site-packages/google/genai/_api_client.py:1103, in BaseApiClient._request_once(self, http_request, stream)
   1095 else:
   1096   response = self._httpx_client.request(
   1097       method=http_request.method,
   1098       url=http_request.url,
   (...)
   1101       timeout=http_request.timeout,
   1102   )
-> 1103   errors.APIError.raise_for_response(response)
   1104   return HttpResponse(
   1105       response.headers, response if stream else [response.text]
   1106   )

File /opt/conda/lib/python3.11/site-packages/google/genai/errors.py:108, in APIError.raise_for_response(cls, response)
    106 status_code = response.status_code
    107 if 400 <= status_code < 500:
--> 108   raise ClientError(status_code, response_json, response)
    109 elif 500 <= status_code < 600:
    110   raise ServerError(status_code, response_json, response)

ClientError: 429 RESOURCE_EXHAUSTED. {'error': {'code': 429, 'message': 'You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.', 'status': 'RESOURCE_EXHAUSTED', 'details': [{'@type': 'type.googleapis.com/google.rpc.Help', 'links': [{'description': 'Learn more about Gemini API quotas', 'url': 'https://ai.google.dev/gemini-api/docs/rate-limits'}]}]}}

I have checked multiple times, and I am confident I am using the right API key and that we are in Tier 1. I haven’t used the API for anything else, and the fact that it says 429 Too Many Requests is also weird as it just happens on the very first request…

Any help or pointers is appreciated. We are trying to adopt the gemini-embedding-001 model in our pipeline, but not being able to use the batch api and the false flags around rate limits is a deal breaker for us.

Scott_Vaillancourt · September 24, 2025, 2:33pm

I was experiencing the same issue in that my requests would receive an immediate 429 error when I tried to create a batch embedding job. What I finally discovered was that I needed to enable the ‘Batch API’ for my project within the Cloud Console. Not sure if this is what you’re running into, but this solved my problem straight away.

Lalit_Kumar · September 26, 2025, 7:15am

Hello,

Welcome to the Forum!!

Could you please share a few additional details? Specifically:

Are you facing this issue only with the Batch API, or also with regular API calls?
How frequently do you observe this issue?
Does it occur at specific times, or is it random?

These details will help us analyze the issue more effectively and provide a more accurate solution.

Sankalp_Singh · September 26, 2025, 7:18am

yes same mee to 429 in tier one i guess its due to new model flash latest does gemini api changed??

Lalit_Kumar · September 29, 2025, 11:13am

Hello,

Welcome to the Forum!

May I kindly request you to share the complete error code along with any additional details you can provide?

Sushant_Karki · September 29, 2025, 1:56pm

Hi Lalit,

Are you facing this issue only with the Batch API, or also with regular API calls?

It seems to happen with the regular embedding api as well. Here is a sample code I ran which failed after generating a few embeddings:

import os
import json
from google import genai
from tqdm import trange

GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY")
client = genai.Client(api_key=GEMINI_API_KEY)

batches = ["batch-1.json", "batch-2.json", "batch-3.json"]

for filename in batches:
    with open(filename, 'r') as file:
        data = json.load(file)
        count_emb = len(data["data"]["embeddings_to_generate"])
        print(f"Loaded {filename}")
        print(f"Number of data points: {count_emb}")
        rows = data["data"]["embeddings_to_generate"]
        batch_size = 100
        for i in trange(0, len(rows), batch_size):
            batch = rows[i:i + batch_size]
            client.models.embed_content(
                model="gemini-embedding-001",
                contents=[r["text"] for r in batch]
            )

And here are the results:

Loaded batch-1.json
Number of data points: 1937
100%|██████████| 20/20 [00:24<00:00,  1.23s/it]
Loaded batch-2.json
Number of data points: 1913
 50%|█████     | 10/20 [00:11<00:11,  1.16s/it]
---------------------------------------------------------------------------
ClientError                               Traceback (most recent call last)
Cell In[2], line 21
     19 for i in trange(0, len(rows), batch_size):
     20     batch = rows[i:i + batch_size]
---> 21     client.models.embed_content(
     22         model="gemini-embedding-001",
     23         contents=[r["text"] for r in batch]
     24     )

File /opt/conda/lib/python3.11/site-packages/google/genai/models.py:5559, in Models.embed_content(self, model, contents, config)
   5556 request_dict = _common.convert_to_dict(request_dict)
   5557 request_dict = _common.encode_unserializable_types(request_dict)
-> 5559 response = self._api_client.request(
   5560     'post', path, request_dict, http_options
   5561 )
   5563 response_dict = '' if not response.body else json.loads(response.body)
   5565 if self._api_client.vertexai:

File /opt/conda/lib/python3.11/site-packages/google/genai/_api_client.py:1290, in BaseApiClient.request(self, http_method, path, request_dict, http_options)
   1280 def request(
   1281     self,
   1282     http_method: str,
   (...)
   1285     http_options: Optional[HttpOptionsOrDict] = None,
   1286 ) -> SdkHttpResponse:
   1287   http_request = self._build_request(
   1288       http_method, path, request_dict, http_options
   1289   )
-> 1290   response = self._request(http_request, http_options, stream=False)
   1291   response_body = (
   1292       response.response_stream[0] if response.response_stream else ''
   1293   )
   1294   return SdkHttpResponse(headers=response.headers, body=response_body)

File /opt/conda/lib/python3.11/site-packages/google/genai/_api_client.py:1126, in BaseApiClient._request(self, http_request, http_options, stream)
   1123     retry = tenacity.Retrying(**retry_kwargs)
   1124     return retry(self._request_once, http_request, stream)  # type: ignore[no-any-return]
-> 1126 return self._retry(self._request_once, http_request, stream)

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:477, in Retrying.__call__(self, fn, *args, **kwargs)
    475 retry_state = RetryCallState(retry_object=self, fn=fn, args=args, kwargs=kwargs)
    476 while True:
--> 477     do = self.iter(retry_state=retry_state)
    478     if isinstance(do, DoAttempt):
    479         try:

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:378, in BaseRetrying.iter(self, retry_state)
    376 result = None
    377 for action in self.iter_state.actions:
--> 378     result = action(retry_state)
    379 return result

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:420, in BaseRetrying._post_stop_check_actions.<locals>.exc_check(rs)
    418 retry_exc = self.retry_error_cls(fut)
    419 if self.reraise:
--> 420     raise retry_exc.reraise()
    421 raise retry_exc from fut.exception()

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:187, in RetryError.reraise(self)
    185 def reraise(self) -> t.NoReturn:
    186     if self.last_attempt.failed:
--> 187         raise self.last_attempt.result()
    188     raise self

File /opt/conda/lib/python3.11/concurrent/futures/_base.py:449, in Future.result(self, timeout)
    447     raise CancelledError()
    448 elif self._state == FINISHED:
--> 449     return self.__get_result()
    451 self._condition.wait(timeout)
    453 if self._state in [CANCELLED, CANCELLED_AND_NOTIFIED]:

File /opt/conda/lib/python3.11/concurrent/futures/_base.py:401, in Future.__get_result(self)
    399 if self._exception:
    400     try:
--> 401         raise self._exception
    402     finally:
    403         # Break a reference cycle with the exception in self._exception
    404         self = None

File /opt/conda/lib/python3.11/site-packages/tenacity/__init__.py:480, in Retrying.__call__(self, fn, *args, **kwargs)
    478 if isinstance(do, DoAttempt):
    479     try:
--> 480         result = fn(*args, **kwargs)
    481     except BaseException:  # noqa: B902
    482         retry_state.set_exception(sys.exc_info())  # type: ignore[arg-type]

File /opt/conda/lib/python3.11/site-packages/google/genai/_api_client.py:1103, in BaseApiClient._request_once(self, http_request, stream)
   1095 else:
   1096   response = self._httpx_client.request(
   1097       method=http_request.method,
   1098       url=http_request.url,
   (...)
   1101       timeout=http_request.timeout,
   1102   )
-> 1103   errors.APIError.raise_for_response(response)
   1104   return HttpResponse(
   1105       response.headers, response if stream else [response.text]
   1106   )

File /opt/conda/lib/python3.11/site-packages/google/genai/errors.py:108, in APIError.raise_for_response(cls, response)
    106 status_code = response.status_code
    107 if 400 <= status_code < 500:
--> 108   raise ClientError(status_code, response_json, response)
    109 elif 500 <= status_code < 600:
    110   raise ServerError(status_code, response_json, response)

ClientError: 429 RESOURCE_EXHAUSTED. {'error': {'code': 429, 'message': 'You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\n* Quota exceeded for metric: generativelanguage.googleapis.com/embed_content_paid_tier_requests, limit: 0', 'status': 'RESOURCE_EXHAUSTED', 'details': [{'@type': 'type.googleapis.com/google.rpc.QuotaFailure', 'violations': [{'quotaMetric': 'generativelanguage.googleapis.com/embed_content_paid_tier_requests', 'quotaId': 'EmbedContentPerMinutePerProjectPerUserPerModel-PaidTier'}]}, {'@type': 'type.googleapis.com/google.rpc.Help', 'links': [{'description': 'Learn more about Gemini API quotas', 'url': 'https://ai.google.dev/gemini-api/docs/rate-limits'}]}]}}

How frequently do you observe this issue?

It happens every time I try to process my batch. For example, when using the same 3 files as the code above (i.e. batch-1.json, batch-2.json, batch-3.json) with the batch api, I have never been able to create batch runs for all 3 files, sometimes one of them succeeds, but then fails, sometimes the first two succeeds then hits limit in the third.

Does it occur at specific times, or is it random?

It doesn’t seem to have to do anything with the time, it happens throughout the day.

MXGordon · October 6, 2025, 1:47am

I’m doing the exact same thing and also getting the exact same error every time. 30 Requests, 100 texts per request, 48120 tokens total. And I’m receiving:

google.genai.errors.ClientError: 429 RESOURCE_EXHAUSTED. {‘error’: {‘code’: 429, ‘message’: ‘You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\n* Quota exceeded for metric: generativelanguage.googleapis.com/embed_content_paid_tier_requests, limit: 0’, ‘status’: ‘RESOURCE_EXHAUSTED’, ‘details’: \[{‘@type’: ‘type.googleapis.com/google.rpc.QuotaFailure’, ‘violations’: \[{‘quotaMetric’: ‘generativelanguage.googleapis.com/embed_content_paid_tier_requests’, ‘quotaId’: ‘EmbedContentPerMinutePerProjectPerUserPerModel-PaidTier’}\]}, {‘@type’: ‘type.googleapis.com/google.rpc.Help’, ‘links’: \[{‘description’: ‘Learn more about Gemini API quotas’, ‘url’: ‘https://ai.google.dev/gemini-api/docs/rate-limits’}\]}\]}}

Have you made any progress?

Sushant_Karki · October 15, 2025, 5:14pm

@Lalit_Kumar Any updates on this?

Megan_Soule · October 23, 2025, 3:13pm

@Lalit_Kumar Do you have an update for @Sushant_Karki and I on this? This has become a critical blocker for us.

Vinga_Luksaite · October 23, 2025, 8:04pm

Yes! I do have an update - fresh from the battle with this 429!

Okay so first, check the basic setup, you do need to have enabled: Generative Language API, Batch API, Compute Engine API, Cloud Logging API.

Next, more context, as my own case may be different from what some of you are encountering. I was hitting my limits with gemini-embedding-001, with a strange error message of zero rate allowance:

``
Job did not complete successfully
Status: failed
Details: An unexpected error occurred during macro indexing: 429 You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: link. To monitor your current usage, head to: link.

Quota exceeded for metric: link/embed_content_paid_tier_requests, limit: 0 [violations {
quota_metric: “link/embed_content_paid_tier_requests”
quota_id: “EmbedContentPerMinutePerProjectPerUserPerModel-PaidTier”
}
```

‘limit 0’ - was what was really confusing.

In my case, embedding generations, some basic checks again:

# Gemini API limits 
MAX_BATCH_SIZE = 100 # Maximum texts per batch
MAX_BATCH_TOKENS = 20000 # Maximum tokens per batch

Plus - in my case, where there were also failures was the 2048 token limit is per individual text in a batch, not just batch total.

Since I was chunking by paragraph, and it was code documentation, turns out some of the paragraphs were exceeding this token length.
I had to make sure the offending paragraphs fitted within token limit, so no failures or silent truncation of text occurred.

And the final thing that was still causing me problems after all that got fixed, was that since I started with Ollama, I was heavily over-optimised for large text chunking. I was creating multiple workers that were completely overwhelming the API in a way that it was causing some generic error (of rate limit zero) which made no sense.

Even after heavily de-optimising the optimisation pixies, requests were still failing with API rate limit (per minute) after some time of generation.
The final fix was to implement a ‘wait’ after hitting the API rate error. The script pauses, waits 30 secs plus 30 secs more if not yet there, and then resumes generating embeddings.

Topic		Replies	Views
Understanding Rate Limits - Tier 3 Account Documentation ai-studio , api , models , gemini , billing	0	34	December 19, 2025
500 INTERNAL SERVER ERRROR when processing more that one file Gemini API bug , api , gemini-2-5	1	69	October 3, 2025
RESOURCE_EXHAUSTED when use gemini-1.5-pro-002 Gemini API gemini-15	8	1600	October 2, 2024
429 Too Many Requests on Vertex AI API generateContent (Gemini 2.5 Pro) Gemini API ai-studio , vertexai , vertex-ai , gemini_25_pro	8	336	December 29, 2025
Gemini Batch Requests are Running Very Slowly and Timing Out Gemini API bug , api	1	195	September 12, 2025

Using Batch API to generate embeddings but getting 429 RESOURCE_EXHAUSTED errors in Paid Tier 1

Related topics