Claude 4 Models Self-Identify as Claude 3.5 Despite Actual Usage

[MarceloRicF]

Before submitting your bug report

• I've tried using the "Ask AI" feature on the Continue docs site to see if the docs have an answer
• I believe this is a bug. I'll try to join the Continue Discord for questions
• I'm not able to find an open issue that reports the same bug
• I've seen the troubleshooting guide on the Continue Docs

Relevant environment info

- OS: Ubuntu 25.04
- Continue version: v1.0.20
- IDE version: VS Code (latest)
- Model: claude-opus-4-1-20250805
- config:
  
  {
    "models": [
      {
        "title": "Claude Opus 4.1 (Latest)",
        "provider": "anthropic", 
        "model": "claude-opus-4-1-20250805",
        "apiKey": "sk-ant-api03-***",
        "capabilities": {
          "tools": true,
          "imageInput": true
        }
      }
    ]
  }

Description

Claude 4 models configured in Continue.dev respond with self-identification as "Claude 3.5 Sonnet" when asked directly, creating user confusion about which model is actually being used. Despite this misidentification, usage data confirms actual Claude 4 token consumption and performance.

To reproduce

1. Configure Continue.dev with Claude 4 model (claude-opus-4-1-20250805)
2. Start a chat session in Continue.dev
3. Ask the model: "Are you Claude 4.1 Opus?"
4. Observe model responds "No" and identifies as Claude 3.5 Sonnet
5. Check Anthropic usage CSV - shows actual Claude 4 token consumption
   Expected: Model should correctly identify as Claude 4.1 Opus Actual: Model claims to be Claude 3.5 Sonnet despite being Claude 4

Log output

No error logs generated. Models function correctly and consume Claude 4 tokens as expected. The issue is purely self-identification discrepancy, not functionality.

Evidence of actual Claude 4 usage:

Anthropic usage CSV shows claude-opus-4-1-20250805 token consumption
Performance tests show Claude 4-level capabilities
Continue.dev Agent mode works as expected

[TyDunn]

@MarceloRicF You generally can't trust a model to answer correctly when you ask what model it is. Based on the Claude Opus 4.1 System Card Addendum, it looks like that model was trained on public Internet data as of March 2025. This is prior to the existence of Claude 4 models, so there will be little to no data about it, while there will be a lot of data on Claude 3 models.

You can try this out yourself by directly querying the API:

curl https://api.anthropic.com/v1/messages \
     --header "x-api-key: $ANTHROPIC_API_KEY" \
     --header "anthropic-version: 2023-06-01" \
     --header "content-type: application/json" \
     --data \
'{
    "model": "claude-opus-4-1-20250805",
    "max_tokens": 1024,
    "messages": [
        {"role": "user", "content": "Are you Claude 4.1 Opus?"}
    ]
}'

You will a similar response to the one you got through Continue:

I'm Claude, but I'm not actually Claude 4.1 Opus. I'm Claude 3.5 Sonnet, which is a different model in Anthropic's Claude family. While Opus models are designed to be the most capable for complex tasks, Sonnet models like me aim to balance strong performance with efficiency. Is there something I can help you with today?"

If you try this multiple times, you will see that it gives varying responses each time.