[BUG] Multi-model response feature fails to honor per-model parameter overrides

Steps to Reproduce:

1. Enable multi-model response as per the feature here: https://typingmind.canny.io/feature-requests/p/multi-model-responses-at-the-same-time.

2. Add two models to run in parallel (e.g., Claude 3.7 Sonnet and Gemini 2.5).

3. Set individual parameters for each model using the "Override Global Model Parameters" option:

• Set Claude 3.7 temperature to default (which should be 1 for extended thinking)

• Set Gemini temperature to 0.2

1. Run a prompt with both models.

Expected Behavior:

Each model should use the specific parameters set in its individual override configuration. Claude should run with temperature 1 (default for thinking), and Gemini should run with temperature 0.2. Token limits and other parameters should also be respected per model.

Actual Behavior:

The system appears to apply the first model's parameter settings globally to all models in the multi-model response. As a result:

• Claude 3.7 fails with the error: "temperature may only be set to 1 when thinking is enabled."

• GPT-4.1 fails with the error: "max tokens is too large: 128000. This model supports at most 32768 completion tokens."

• Gemini outputs using incorrect temperature settings, making comparisons inconsistent and invalid.

Screenshots:

• GPT-4.1 token limit error shown in multi-model response

• Claude 3.7 temperature error in multi-model usage