- reasoning_effort was forwarded only to IDs matching gpt-5 or the o-series, so
GPT-6 models silently dropped the configured effort; match GPT-5 and later,
with a version boundary so unrelated IDs do not match
- Gemini Pro, 3.8+ and the -latest aliases reject thinking_level "minimal"
with a 400; send it only to numbered Flash models before 3.8 and map it to
"low" elsewhere, since aliases move between generations
Trim each provider to current-generation models and drop the special-casing
they required:
- OpenAI: remove gpt-4.1 (deprecated; the only non-reasoning model).
- Anthropic: remove Claude Sonnet 4.5 (legacy; the only Sonnet that 400s on effort).
- Google: remove the Gemini 2.5 line (superseded by 3.x).
- Gemini client: drop the integer thinking_budget mapping; 3.x takes the string
thinking_level directly.
Effort/reasoning gates stay as defense in depth for custom model IDs. All kept
IDs verified against live APIs.
Yijia-Xiao