Google: Gemini 3.7 Flash (batch)

Gemini 3.7 Flash combines multimodal reasoning with coding and multi-step agent work. Google’s 3.7 profile emphasizes design-mock fidelity and agentic video understanding, with low, medium and high reasoning. It accepts text and media, returns text, and has documented 1,048,576/65,536-token input/output limits.

Ailoot Ultra · Batch POST /v1/batches google/gemini-3.7-flash:batch

About the model

This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.

Gemini 3.7 Flash is Google’s named multimodal reasoning profile for coding, agentic workflows and multi-step execution. It builds on Gemini 3.6 Flash, with algorithmic changes to reasoning and support for agentic video understanding. Text, images, audio, video and PDF can form the source material for a text result. Useful deliverables include a focused code revision, a comparison between a design mock and an implementation, or a source-grounded analysis of events in a recording.

The 3.7 developer guide highlights stronger design parity in desktop and web work. For that task, provide the intended mock, the relevant source and the observable discrepancy. Ask for a comparison against concrete elements such as spacing, component states, text hierarchy or interaction behaviour. For a design comparison, request a list of mismatches between a settings-panel mock and its existing implementation, with each proposed correction tied to the screenshot or source. The result should identify what needs changing without silently extending the design beyond the supplied requirements.

For a coding workflow, give the goal, existing behaviour, relevant files and constraints. Google positions 3.7 for resolving software issues and reducing failed agent loops, but a useful review still checks the actual modification and validation result. Keep the sequence of tool observations in the conversation so a later step does not start from an invented account of the codebase. The final result can distinguish a proposed change from one confirmed through the connected application’s tools.

The documented thinking levels are low, medium and high, with medium as the default. Low supports quicker bounded answers, while high is intended for difficult reasoning and extended tool work. Gemini 3.7 Flash does not support minimal thinking. A short analysis of one interface mismatch and a deeper multi-file investigation therefore call for different effort choices. Define the final artifact and the conditions it must satisfy before requesting a longer sequence of reasoning and actions.

The exact model page lists function calling, code execution, file search, search grounding, Maps grounding and URL context. Computer Use is marked as preview, and structured outputs can define a result record. A function request is a proposed operation for the connected application to execute and return. For an agent task, give the tool a clear purpose and retain the real result before asking the model to continue. For a document extraction, define the fields, acceptable values and how missing information should be represented.

This edition supports agentic video understanding. Supply the relevant recording and state the event or relationship to examine. A practical brief can ask which visible action explains a state change, or request an ordered description of events with evidence from the clip. Images and PDF pages can similarly support a visual or document comparison. The output remains text: interpretation, extracted data, code or a written explanation. Generated video, artwork or speech belongs to separate model profiles.

Google’s exact 3.7 API table specifies 1,048,576 input tokens and 65,536 output tokens, while its native card describes up to 1M context and 64K text output. Include instructions, prior turns and supplied media in the input preparation. Batch processing and caching are listed for the producer profile, alongside Flex and Priority options. For a collection of independent requests, retain source identifiers and a consistent result format so each answer can be checked against its own input.

Possible limitations include hallucinations and occasional slowness or timeouts. Its knowledge cutoff is March 2026, with some domains limited to January 2025. Provide current source material when a conclusion depends on more recent facts. The edition’s useful distinction is its documented focus on software issues, design adherence and video reasoning; inspect those results against the actual mock, code or media rather than replacing a named historical profile with newer features.

Ailoot Ultra · Batch POST /v1/batches google/gemini-3.7-flash:batch

google/gemini-3.7-flash:batch

Google /google/gemini-3.7-flash:batch

Overview

Accepted inputs: Text
Results: Text

Available in Ailoot · inputs & methods

Ailoot Ultra · Batch POST /v1/batches

Input types: text
Output types: text
Output format: inline results[].response.body JSON

Field	Type	Requirement	Values & limits	Default
Field`endpoint`	Type`string`	RequirementRequired	Values & limits /v1/chat/completions	Default—
Field`model`	Type`string`	RequirementRequired	Values & limits google/gemini-3.7-flash:batch	Default—
Field`requests`	Type`batch_item[]`	RequirementRequired	Values & limits Minimum items: 1	Default—
Field`requests[].custom_id`	Type`string`	RequirementRequired	Values & limits	Default—
Field`requests[].body.messages`	Type`message[]`	RequirementRequired	Values & limits	Default—
Field`completion_window`	Type`string`	RequirementOptional	Values & limits 24h	Default—
Field`provider`	Type`object\|null`	RequirementOptional	Values & limits	Default—
Field`requests[].body.model`	Type`string`	RequirementOptional	Values & limits google/gemini-3.7-flash:batch, google/gemini-3.7-flash	Default—
Field`provider.only`	Type`string[]`	RequirementOptional	Values & limits google-vertex/global	Default—
Field`provider.allow_fallbacks`	Type`boolean\|null`	RequirementOptional	Values & limits false, null	Default—

Method parameters and lifecycle

Parameter	Contract
Parameter`content_type`	Contractapplication/json
Parameter`required`	Contractendpoint, model, requests
Parameter`material_types`	Contracttext
Parameter`representation`	Contractinline requests[]; each item has custom_id and body
Parameter`request_endpoint_profile`	Contract/v1/chat/completions
Parameter`file_material`	ContractCaller may read actual UTF-8 JSONL locally into requests[]. No upload purpose, input_file_id or file reference is used. Missing/empty customer material stops preparation before POST.
Parameter`serialization_order`	Contractendpoint, model, provider (if supplied), completion_window (if supplied), requests last
Parameter`optional_controls`	Contractprovider.only, provider.allow_fallbacks, completion_window
Parameter`schema_scope`	ContractText Chat Completions bodies in the documented inline Batch envelope.

Lifecycle: asynchronous

GET /v1/batches/{operation_id} · actual original create response.id; same user/key/group

Pricing

Tariff references do not add request options. Supported parameters are listed for the selected method.

Fast mode: Not specified When fast mode is selected
Flex mode: Not specified When flex mode is selected
Maximum reasoning effort: Not specified When maximum reasoning effort uses the default adjustment
Explicit cache pricing: Not enabled When explicit cache is requested

Ailoot rates

Prices by configuration

Match each charge to the conditions shown. Use one rate for each set of alternative configurations.

Configuration	Rate schedule	Context tokens, more than	Context tokens, up to and including	Charge
ConfigurationBase token rates	Rate scheduleBase rate	Context tokens, more than—	Context tokens, up to and includingNo upper bound	Charge Input · Input tokens $0.00075 / 1K tokens Output · Output tokens $0.00375 / 1K tokens Input · Cache read $0.000075 / 1K tokens Input · Cache write $0.0000833333333333334 / 1K tokens Input · Cache write, 1 hour — / 1K tokens
ConfigurationRequest-length tier 1	Rate schedule—	Context tokens, more than0	Context tokens, up to and includingNo upper bound	Charge Input · Input tokens $0.00075 / 1K tokens Output · Output tokens $0.00375 / 1K tokens Input · Cache read $0.000075 / 1K tokens Input · Cache write $0.0000833333333333334 / 1K tokens Input · Cache write, 1 hour — / 1K tokens

Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →

Example requests

Ailoot Ultra · Batch

curl · retained fixturePython HTTP · actual customer JSONLPython · save every completed result

set -eu
: "${AILOOT_API_KEY:?Set your Ailoot API key}"
python3 - <<'PY_REQUEST'
import json
from pathlib import Path
body=json.loads('{"endpoint":"/v1/chat/completions","model":"google/gemini-3.7-flash:batch","requests":[{"custom_id":"req-0001","body":{"messages":[{"role":"user","content":"Hello!"}],"max_tokens":16}}]}')
Path("batch-request.json").write_text(json.dumps(body,ensure_ascii=False),encoding="utf-8")
PY_REQUEST
curl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/batches \
  -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \
  --data-binary @batch-request.json -D batch-response-headers.txt -o batch-create-response.json

import json, os
from pathlib import Path
from urllib.request import Request, urlopen

def prepare_requests(customer_path):
    path = Path(customer_path)
    if not path.is_file(): raise ValueError("Customer JSONL file is missing")
    items = [json.loads(line) for line in path.read_text(encoding="utf-8").splitlines() if line.strip()]
    if not items: raise ValueError("Customer JSONL contains no requests")
    ids = set()
    for item in items:
        if set(item) != {"custom_id", "body"}: raise ValueError("Each item needs custom_id and body")
        cid = item["custom_id"]
        if not isinstance(cid, str) or not cid.strip() or cid in ids: raise ValueError("custom_id must be nonempty and unique")
        ids.add(cid)
        if not isinstance(item["body"], dict) or not item["body"].get("messages"): raise ValueError("Supply actual text messages")
        supplied_model = item["body"].get("model")
        if supplied_model is not None and supplied_model not in ['google/gemini-3.7-flash:batch', 'google/gemini-3.7-flash']: raise ValueError("body.model must be an exact documented caller or wire ID")
    return {"endpoint":"/v1/chat/completions", "model":"google/gemini-3.7-flash:batch", "requests":items}

if __name__ == "__main__":
    material = os.environ.get("AILOOT_BATCH_REQUESTS_JSONL", "")
    if not material: raise SystemExit("Set AILOOT_BATCH_REQUESTS_JSONL to your actual customer file")
    body = prepare_requests(material)
    key = os.environ.get("AILOOT_API_KEY", "")
    if not key: raise SystemExit("Set AILOOT_API_KEY")
    request = Request("https://api.ailoot.ai/v1/batches", data=json.dumps(body, ensure_ascii=False).encode("utf-8"), headers={"Authorization":"Bearer " + key, "Content-Type":"application/json"}, method="POST")
    with urlopen(request) as response:
        raw = response.read()
        Path("batch-create-response.json").write_bytes(raw)
        print(raw.decode("utf-8"))

import json
from pathlib import Path
raw = Path("batch-status-response.json").read_bytes()
response = json.loads(raw)
if response.get("status") != "completed":
    raise SystemExit("Keep the original operation; read its status without resubmitting")
results = response.get("results")
if not isinstance(results, list) or not results:
    raise SystemExit("Preserve the original response; completed results are missing")
Path("batch-results.jsonl").write_text("".join(json.dumps(item, ensure_ascii=False) + "\n" for item in results), encoding="utf-8")
# Preserve batch-status-response.json with aggregate usage and every result/error.

Receive and save results

Ailoot Ultra · Batch

inline results[].response.body JSON

GET /v1/batches/{operation_id}

Check job status

curl · original operation status

set -eu
: "${AILOOT_API_KEY:?Set your Ailoot API key}"
BATCH_OPERATION_ID="$(python3 - <<'PY_ID'
import json
from pathlib import Path
from urllib.parse import quote
response = json.loads(Path("batch-create-response.json").read_text(encoding="utf-8"))
operation = response.get("id")
if not isinstance(operation, str) or not operation: raise SystemExit("No actual original operation ID; preserve the response and do not resubmit")
print(quote(operation, safe=""))
PY_ID
)"
curl --fail-with-body --silent --show-error "https://api.ailoot.ai/v1/batches/$BATCH_OPERATION_ID" \
  -H "Authorization: Bearer $AILOOT_API_KEY" -o batch-status-response.json

Keep all completed results

Keep every complete returned response, file and content part.

Preserve the full original status JSON, every result/custom_id and complete response or error body. Save completed results[] as JSONL without filtering.

If the original response needs checking

Preserve the original operation ID. 503 result_pending_check has X-Should-Retry:false; never resubmit an unknown original batch.

Response fields

Field	Type
Field`results[].custom_id`	Typestring
Field`results[].response.body.choices[].message.content`	Typestring\|null
Field`results[].error`	Typeobject\|null
Field`usage`	Typeobject\|null
Field`id`	Typestring
Field`model`	Typestring
Field`endpoint`	Typestring
Field`status`	Typestring
Field`request_counts`	Typeobject
Field`results[].id`	Typestring
Field`results[].response.status_code`	Typeinteger
Field`results[].response.request_id`	Typestring

Errors and next steps

Response	Condition	Next step
Response400 · `invalid_request_error`	ConditionInvalid documented inline envelope, duplicate keys, item identity, or configured model capabilities conflict.	Next stepCorrect the indicated fields before any new submission.
Response401 · `authentication_error`	ConditionInvalid API key.	Next stepUse the authorized Ailoot key.
Response403 · `permission_error`	ConditionBatch group unavailable.	Next stepCheck key/group entitlement.
Response422 · `invalid_request_error`	ConditionBatch uses balance billing; subscription creation is rejected.	Next stepUse the authorized balance-billed group.
Response422 · `price_unavailable`	ConditionBatch input cannot be quoted.	Next stepPreserve response; use documented family input.
Response402 · `billing_error`	ConditionBatch reserve unavailable.	Next stepResolve admission before a new request.
Response503 · `price_unavailable`	ConditionBatch method price/native rates unavailable.	Next stepPreserve response; do not substitute prices.
Response503 · `api_error`	ConditionExact supplier route unavailable.	Next stepPreserve response for route repair.
Response503 · `result_pending_check`	ConditionOriginal result requires checking; X-Should-Retry:false.	Next stepKeep original operation_id/id and read its status; never repeat original POST.
Response404 · `not_found`	ConditionBatch status operation absent or owned by another identity.	Next stepUse exact original id and same user/key/group.

Aliases

Model aliases

Ailoot Ultra google/gemini-3.7-flash:batch

Method path aliases

Use the method paths shown above.

Related caller IDs

google/gemini-3.7-flash:batch

FAQ · Google: Gemini 3.7 Flash (batch)

What inputs and results are available?: POST /v1/batches · google/gemini-3.7-flash:batch: text → text
Which fields does the request use?: POST /v1/batches · google/gemini-3.7-flash:batch: endpoint, model, requests, completion_window, provider
What limits apply?: POST /v1/batches · google/gemini-3.7-flash:batch: requests.min_items: 1

Google: Gemini 3.7 Flash (batch)