AWS bedrock: Pegasus v1.5 model card: token limits, endpoints, tiers, region routing
Summary
Adds context window (261,120 tokens) and max output token (98,304) details, adds an In-Region endpoint URL column (N/A) with an example global cross-Region inference endpoint URL, adds a 'Batch' service tier column, documents how the US and global inference profiles route requests, and adds a note that the s3Location.uri bucket Region must match the request Region or invocation fails with a ValidationException.
Security assessment
The change is a model-card/usage update: token limits, endpoint table columns, supported service tiers, cross-Region inference profile routing, and a note that the S3 input bucket must be in the same Region as the request. The Region note has mild data-handling/compliance relevance (it explains where video data is read from and that requests are served across regions via geo/global profiles), but it documents functional behavior rather than introducing or fixing a vulnerability, credential handling, encryption, or authorization change.
Evidence
Replace `s3Location.uri` with a bucket in the same AWS Region that you send the request to — `us-east-1` in this example. The bucket Region must match the request Region even though the model is accessed through a geo or global inference profile, and a bucket in any other Region fails with a `ValidationException`.
Diff
diff --git a/bedrock/latest/userguide/model-card-twelvelabs-pegasus-v1-5.md b/bedrock/latest/userguide/model-card-twelvelabs-pegasus-v1-5.md index 7fb67c0..7bb2b0a 100644 --- a//bedrock/latest/userguide/model-card-twelvelabs-pegasus-v1-5.md +++ b//bedrock/latest/userguide/model-card-twelvelabs-pegasus-v1-5.md @@ -30,0 +31,4 @@ Pegasus v1.5 is TwelveLabs' video-to-text generation model that produces summari + * **Context window:** 261,120 tokens + + * **Max output tokens:** 98,304 tokens + @@ -88,3 +92,5 @@ For the full request parameters, response fields, and request and response examp -**Endpoint** | **Model ID** | **Geo inference ID** | **Global inference ID** ----|---|---|--- -`bedrock-runtime` | `twelvelabs.pegasus-1-5-v1:0` | `us.twelvelabs.pegasus-1-5-v1:0` | `global.twelvelabs.pegasus-1-5-v1:0` +**Endpoint** | **Model ID** | **In-Region endpoint URL** | **Geo inference ID** | **Global inference ID** +---|---|---|---|--- +`bedrock-runtime` | `twelvelabs.pegasus-1-5-v1:0` | N/A | `us.twelvelabs.pegasus-1-5-v1:0` | `global.twelvelabs.pegasus-1-5-v1:0` + +_For example, to use Global cross-Region inference from us-east-1 (N. Virginia), use the endpoint URL "https://bedrock-runtime.us-east-1.amazonaws.com" with the inference profile ID "global.twelvelabs.pegasus-1-5-v1:0"._ @@ -96,3 +102,3 @@ Amazon Bedrock offers multiple service tiers to match your workload requirements -**Standard** | **Priority** | **Flex** | **Reserved** ----|---|---|--- - |  |  |  +**Standard** | **Priority** | **Flex** | **Reserved** | **Batch** +---|---|---|---|--- + |  |  |  |  @@ -107,0 +114,9 @@ Pegasus v1.5 supports Geo Cross-Region inference from the US Regions and Global +The Region you call is the source Region. The inference profile you use determines where the request can be served: + + * The US geographic inference profile (`us.twelvelabs.pegasus-1-5-v1:0`) routes requests to us-east-1, us-east-2, and us-west-2. + + * The global inference profile (`global.twelvelabs.pegasus-1-5-v1:0`) routes requests globally across all supported AWS Regions. + + + + @@ -186,0 +202,4 @@ Invoke API +###### Note + +Replace `s3Location.uri` with a bucket in the same AWS Region that you send the request to — `us-east-1` in this example. The bucket Region must match the request Region even though the model is accessed through a geo or global inference profile, and a bucket in any other Region fails with a `ValidationException`. For the full parameter reference, see [TwelveLabs Pegasus 1.5](./model-parameters-pegasus-1-5.html). +