[FEATURE] Allow model alias mapping for Bedrock Inference Profiles in Model List

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Summary

When using AWS Bedrock Inference Profiles in Cowork 3P's Model List, there is no way to assign a friendly display name or map an inference profile ARN to a standard model alias (sonnet, opus, haiku). This means inference profiles can't be set as the default model for a given tier the way standard model IDs can in Claude Code.

###Context

• Provider: AWS Bedrock via SSO / Identity Center
• Feature: Cowork 3P Model List configuration
• Org requirement: All requests must route through Bedrock Inference Profiles (not direct model IDs)

Problem

The Model List currently accepts Bedrock Inference Profile ARNs (e.g., arn:aws:bedrock:us-east-2:188572534221:application-inference-profile/sd030ajk), but:

1. No display name / alias: The full ARN is shown in the model picker, which is unwieldy and confusing for end users.
2. No model tier mapping: There's no way to tell Cowork 3P that a given inference profile is the org's sonnet, opus, or haiku model. In Claude Code, you can set aliases like sonnet → a specific model ID, but inference profiles don't have this capability in the Cowork 3P UI.
3. Can't set as default: Because inference profiles can't be mapped to a tier, they can't be set as the "first entry" default in a meaningful way — the model picker just shows a raw ARN.

Proposed Solution

The Model List UI should allow administrators to:

1. Set a display name for each model entry (e.g., "Opus (Production)" instead of the raw ARN).
2. Map an inference profile to a model alias (sonnet, opus, haiku) so that Cowork 3P treats it as that tier's default model. This is how Claude Code handles model configuration and the same UX should apply to inference profiles.
3. Set the default model per tier, so when a user selects "opus" in the model picker, the request routes through the configured inference profile ARN under the hood.

Alternative Solutions

Users must manually select the full ARN from the picker, which is error-prone and doesn't integrate with any tier-based model selection logic.

Priority

Critical - Blocking my work

Feature Category

Other

Use Case Example

_No response_

Additional Context

Environment

• Claude Desktop with Cowork 3P enabled
• AWS Bedrock provider with SSO / Identity Center authentication
• Inference Profiles configured at the AWS account level