AWS bedrock: Claude Haiku 4.5 model card: alt text cleanup and watermarking row
Summary
Replaces verbose image alt text (e.g. 'Green circle with white checkmark icon.') with concise 'supported'/'not-supported' labels across modality and feature-support tables, adds a 'Watermarking' feature row (not supported) to the model/tool support lists, and removes the note stating that structured outputs are unsupported when invoking Claude Haiku 4.5 through the India Geographic cross-Region inference profile.
Security assessment
The diff is documentation hygiene: image alt text is shortened to 'supported'/'not-supported' while the underlying icons and support statuses are unchanged, and a Watermarking capability row is added as not supported. No vulnerability, credential, permission, or exposure change is described; the only functional-narrative edit is deletion of a regional feature-availability caveat for structured outputs, which is a compatibility note rather than a security control. The pre-existing support entries for 'Guardrails' and 'Abuse detection' are merely re-rendered with new alt text, not newly documented as security features.
Evidence
+ *  [Watermarking](watermarking.html)
Diff
diff --git a/bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.md b/bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.md index 0747a54..d3db56d 100644 --- a//bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.md +++ b//bedrock/latest/userguide/model-card-anthropic-claude-haiku-4-5.md @@ -46,5 +46,5 @@ Claude Haiku 4.5 is Anthropic's lightweight model optimized for speed and effici - Audio|  Embedding - Image|  Image - Speech|  Speech - Text|  Text - Video|  Video + Audio|  Embedding + Image|  Image + Speech|  Speech + Text|  Text + Video|  Video @@ -88,13 +88,13 @@ Whenever possible, we recommend using the `bedrock-runtime` endpoint for new app - *  [Response streaming](/bedrock/latest/APIReference/API_runtime_InvokeModelWithResponseStream.html) - *  [Abuse detection](abuse-detection.html) - *  [Guardrails](guardrails.html) - *  [Prompt optimization](prompt-management-optimize.html) - *  [Count tokens](count-tokens.html) - *  [Knowledge base](knowledge-base.html) - *  [Model evaluation](evaluation.html) - *  [Prompt management](prompt-management.html) - *  [Flows](flows.html) - *  [Agents](agents.html) - *  [Structured outputs](./structured-output.html) - *  [Client-side tool calling](tool-use.html) - *  [Computer use](computer-use.html) + *  [Response streaming](/bedrock/latest/APIReference/API_runtime_InvokeModelWithResponseStream.html) + *  [Abuse detection](abuse-detection.html) + *  [Guardrails](guardrails.html) + *  [Prompt optimization](prompt-management-optimize.html) + *  [Count tokens](count-tokens.html) + *  [Knowledge base](knowledge-base.html) + *  [Model evaluation](evaluation.html) + *  [Prompt management](prompt-management.html) + *  [Flows](flows.html) + *  [Agents](agents.html) + *  [Structured outputs](./structured-output.html) + *  [Client-side tool calling](tool-use.html) + *  [Computer use](computer-use.html) @@ -104 +104,2 @@ Whenever possible, we recommend using the `bedrock-runtime` endpoint for new app - *  [Intelligent prompt routing](prompt-routing.html) + *  [Intelligent prompt routing](prompt-routing.html) + *  [Watermarking](watermarking.html) @@ -113,5 +114,5 @@ Whenever possible, we recommend using the `bedrock-runtime` endpoint for new app - *  [Response streaming](/bedrock/latest/APIReference/API_runtime_InvokeModelWithResponseStream.html) - *  [Abuse detection](abuse-detection.html) - *  [Count tokens](count-tokens.html) - *  [Computer use](computer-use.html) - *  [Structured outputs](./structured-output.html) + *  [Response streaming](/bedrock/latest/APIReference/API_runtime_InvokeModelWithResponseStream.html) + *  [Abuse detection](abuse-detection.html) + *  [Count tokens](count-tokens.html) + *  [Computer use](computer-use.html) + *  [Structured outputs](./structured-output.html) @@ -121,8 +122,9 @@ Whenever possible, we recommend using the `bedrock-runtime` endpoint for new app - *  [Guardrails](guardrails.html) - *  [Prompt optimization](prompt-management-optimize.html) - *  [Knowledge base](knowledge-base.html) - *  [Model evaluation](evaluation.html) - *  [Prompt management](prompt-management.html) - *  [Flows](flows.html) - *  [Agents](agents.html) - *  [Intelligent prompt routing](prompt-routing.html) + *  [Guardrails](guardrails.html) + *  [Prompt optimization](prompt-management-optimize.html) + *  [Knowledge base](knowledge-base.html) + *  [Model evaluation](evaluation.html) + *  [Prompt management](prompt-management.html) + *  [Flows](flows.html) + *  [Agents](agents.html) + *  [Intelligent prompt routing](prompt-routing.html) + *  [Watermarking](watermarking.html) @@ -132,4 +133,0 @@ Whenever possible, we recommend using the `bedrock-runtime` endpoint for new app -###### Important - -Structured outputs aren't currently supported when you invoke Claude Haiku 4.5 through the India Geographic cross-Region inference profile (`in.anthropic.claude-haiku-4-5-20251001-v1:0`). To use structured outputs with this model, call it through an inference profile or endpoint that supports the feature. - @@ -185 +183 @@ Amazon Bedrock offers multiple service tiers to match your workload requirements - |  |  |  + |  |  | 