The same guide states 250 texts / 20,000 tokens per request in its API limits section, but five texts in the REST instructions. The per-text 2,048-token limit agrees; the request-count discrepancy is unresolved.Source for Request limits (opens in a new tab)
Access & pricing
Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.
The same guide states 250 texts / 20,000 tokens per request in its API limits section, but five texts in the REST instructions. The per-text 2,048-token limit agrees; the request-count discrepancy is unresolved.Source for Request limits (opens in a new tab)
Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 14 Oct 2026
Category
Usage
Rate
Unit
Embedding pricing: billing unit needs verification
Text input1
$0.000025
1,000 provider billing counts (unit unresolved)
Embedding inference
Embedding output2
$0.00
output response
Rate conditions
Current pricing table labels the denominator as count without defining it. Legacy guidance describes character billing, but the current exact unit needs SKU verification. Do not convert this rate to tokens or use it in a question estimate.
The pricing table explicitly states no output charge. Input is charged separately.
Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 14 Oct 2026
Category
Usage
Rate
Unit
Embedding pricing: billing unit needs verification
Text input1
$0.00002
1,000 provider billing counts (unit unresolved)
Embedding inference
Embedding output2
$0.00
output response
Rate conditions
Current pricing table labels the denominator as count without defining it. Legacy guidance describes character billing, but the current exact unit needs SKU verification. Do not convert this rate to tokens or use it in a question estimate.
The pricing table explicitly states no output charge. Input is charged separately.
Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.
Caching
Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.
01 / INITIAL INPUTProcess & retain
The system processes eligible input and retains reusable state.
02 / LATER REQUESTCheck available state
Reuse depends on matching input, session state and the cache policy.
03 / CACHE HITReuse & generate
Reusing state can reduce repeated work. Any charges follow the billing rules below.
Google Enterprise Agent Platform (online)
text-embedding-005
Not verified
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
How it works
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Minimum reusable input
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Lifetime & refresh
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
How to enable it
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Usage fields
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Billing rules
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Limits & data boundaries
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026
Google Enterprise Agent Platform (batch)
text-embedding-005
Not verified
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
How it works
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Minimum reusable input
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Lifetime & refresh
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
How to enable it
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Usage fields
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Billing rules
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
Limits & data boundaries
Not collected
Caching for this embedding endpoint is not established by the reviewed model docs. Generative Gemini prompt-cache support is not assumed.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026
What the measurements say
No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.
Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.
The record behind the model
Why it is included
Google identifies this proprietary, Google-hosted model separately from open embedding models. No downloadable weights or open-weight license is supplied in the reviewed documentation; closed distribution is inferred, not a claim about unpublished architecture.
