feat(ovhcloud): add Qwen3.6-27B model (262K context, \$0.47/\$3.19 pe…

[elbatoony]

## Summary

When using Goose Desktop 1.41 with the OVHCloud provider, the model `Qwen3.6-27B` is not handled correctly.

The OVH OpenAI-compatible `/v1/models` endpoint reports:

```json
{
  "id": "Qwen3.6-27B",
  "context_length": 262144,
  "max_completion_tokens": 262144
}

However:

• Goose Desktop displays a 128k context window instead of 262k.
• Selecting Qwen3.6-27B from the UI results in an "Invalid params" error.
• The model only works when manually configured in config.yaml.
• Other OVHCloud models (e.g. gpt-oss-120b) are detected correctly and their context window is displayed properly. [github.com], [github.com]

Root Cause

It appears that Qwen3.6-27B is either missing from the model recognition/mapping logic or is not correctly matched by the OVHCloud provider implementation.

As a result, Goose falls back to the default 128k context limit instead of using the context_length value returned by the OVH endpoint. [github.com], [github.com], [github.com]

Proposed Fix

Add support for Qwen3.6-27B in the model mapping logic and use the advertised context window:

"context_length": 262144

This should allow:

• Correct context window display (262k).
• Successful model selection from the UI.
• Consistent behavior with other OVHCloud-hosted models. [github.com], [github.com]

[kursad-k]

This is not a true solution, but you can set the context limit using GOOSE_CONTEXT_LIMIT = 262144.

[elbatoony]

This is not a true solution, but you can set the context limit using GOOSE_CONTEXT_LIMIT = 262144.

Yes i tried but nothing change in desktop interface (still défaut value 128k appear).
LLM and providers change every morning, a better solution will be to manage LLM in a conf file which don't depend of app compilation but I'm not so skilled to purpose a PR like that.