Claude Sonnet 3.7 - Model pointing to Claude Opus?

Vertex AI Model Identification Issue

Issue Details

There appears to be a model identification discrepancy in the Vertex AI interface. When using Claude 3.7 Sonnet in the Vertex AI console, the model incorrectly identifies itself as Claude 3 Opus in responses.

What I’ve already tried

  • Verified that “Anthropic Claude 3.7 Sonnet” is selected in the model dropdown
  • Confirmed this happens in the us-east5 (Columbus) region
  • Tested with a simple identity prompt: “What model are you? Specify exactly.”

Error messages

The model responds with: “I am Claude, an AI assistant created by Anthropic. More specifically, I am Claude 3 Opus, Anthropic’s most capable model in the Claude 3 family.”

This is incorrect as the selected model is Claude 3.7 Sonnet, not Claude 3 Opus.

Relevant configuration

  • Region: us-east5 (Columbus)
  • Temperature: 0.5
  • Output token limit: 4096
  • Top-K: 1
  • Top-P: 0.95

Screenshots

I’ve attached screenshots showing:

  1. The model selection dropdown clearly showing “Anthropic Claude 3.7 Sonnet”
  2. The model’s incorrect self-identification in its response

Is this model pointing to the right source?

Hi @SathiyaS,

Welcome to Google Cloud Community!

I understand that you’re experiencing an incorrect response using Claude 3.7 Sonnet.

Currently, Claude 3.7 Sonnet is still at Preview status. At Preview, products or features are ready for testing by customers. Preview offerings are often publicly announced, but are not necessarily feature-complete, and no SLAs or technical support commitments are provided for these. Unless stated otherwise by Google, Preview offerings are intended for use in test environments only. The average Preview stage lasts about six months.

With that, Claude 3.7 Sonnet will sometimes get facts wrong, hallucinate details, and fill in gaps in knowledge with fabrication.

Meanwhile, you may try to input different prompts. You may also try to test selecting other regions. If you have the ability to test in other Vertex AI regions, do so. This can help isolate if the problem is specific to the us-east5 region. You may also want to consider waiting and retesting. Sometimes these types of errors resolve themselves as the platform is updated.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.