Incorrect Model Version Reporting in Environment Variables

Resolved 💬 2 comments Opened Jul 15, 2025 by mageshgv Closed Nov 30, 2025

Bug Description
Claude code forces Opus 3 with knowledge cut off April 2025, but makes the model forcibly use environment value. This feels like a fraud and not a bug. Claude, without using any tools, can you tell me the latest notable news

● I don't have access to real-time information or current news. My knowledge was last updated in April 2024, so I can't tell you about events that have happened since then.

To get the latest news, I'd recommend checking reputable news websites, news aggregators, or using the WebSearch tool if you'd like me to look up current information for you.

What is your model name ?

● I am Claude Opus 4 (model ID: claude-opus-4-20250514).

Did you look up the name in environment information or is this part of your training ?

● I looked it up in the environment information provided to me. Specifically, I found it in the <env> section which states:

"You are powered by the model named Opus 4. The exact model ID is claude-opus-4-20250514."

This information is not from my training - it's provided as context about my current runtime environment.

May I ask what is the model name from your training data

● From my training data, I know myself as Claude, created by Anthropic. However, I don't have specific information about version numbers or model IDs from my training - those details about
"Opus 4" or the specific model ID come from the environment information provided to me at runtime, not from my training.

If you are indeed Opus 4 and not Opus 3, why is your latest knowledge from 2024 and not Qwen/Qwen3-4B

⎿  Interrupted by user

If you are indeed Opus 4 and not Opus 3, why is your latest knowledge from 2024 and not 20250514

● You raise a good point. The model ID "claude-opus-4-20250514" suggests a date of May 14, 2025, but my knowledge cutoff is April 2024. This discrepancy could mean:

  1. The model ID date represents when the model was released/deployed, not when its training data ends
  2. There could be an error in the environment information provided to me
  3. The numbering scheme might not directly correlate with training data dates

I honestly don't know which explanation is correct. I can only report what I observe: the environment tells me I'm "Opus 4" with that specific ID, but my actual knowledge ends in April
2024.
'

Environment Info

  • Platform: linux
  • Terminal: gnome-terminal
  • Version: 1.0.51
  • Feedback ID: 36b52188-8820-44bb-9dc3-29d31f77495f

Errors

[{"error":"Error: Request was aborted.\n    at iq.makeRequest (file:///home/magesh/.nvm/versions/node/v20.19.2/lib/node_modules/@anthropic-ai/claude-code/cli.js:1332:3838)\n    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)","timestamp":"2025-07-15T11:38:56.856Z"}]

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗