model mismatch. are we being served the model that we ask for? #142
Description
Hi,
Not 100% sure, but it seems as if the model isn't being respected when creating a session. I made a test that you can run, which is here:
https://github.com/EvaLok/copilot-sdk-model-check-test
But basically, this is what's happening..
using client.createSession with the model set is not respecting the model. i asked for gpt-5.1-codex, and i get 4.1. i've tried with other models as well and the problem is consistent, sometimes i get 4o also.
log:
{
"event": "test_start",
"timestamp": "2026-01-23T14:22:05.230Z",
"requestedModel": "gpt-5.1-codex"
}
---
{
"event": "client_started",
"timestamp": "2026-01-23T14:22:05.697Z"
}
---
{
"event": "session_created",
"timestamp": "2026-01-23T14:22:05.748Z",
"configuredModel": "gpt-5.1-codex"
}
---
{
"event": "prompt_sent",
"timestamp": "2026-01-23T14:22:05.748Z",
"prompt": "Please identify yourself. What model are you? What is your exact model name/version? Be specific about your model identifier."
}
---
{
"event": "response_received",
"timestamp": "2026-01-23T14:22:12.110Z",
"messageId": "82a4e545-ba51-4736-bb37-610ceb163f3d",
"response": "I’m GitHub Copilot CLI, a terminal-focused assistant built by GitHub. I’m running on the GPT-4.1 architecture, variant 0.0.392 as shown in the session’s version information."
}
---
{
"event": "test_complete",
"timestamp": "2026-01-23T14:22:12.111Z",
"requestedModel": "gpt-5.1-codex",
"summary": "Check if the model response matches the requested model"
}
---