-
[NA] [FE] fix: mark GPT-5.6 models as reasoning models (#7451)
frostbyte_neo 发布于
2026-07-14 08:27:11 +00:00 The gpt-5.6-luna/sol/terra models were present in the enum and backend
YAML but missing from OPENAI_MODEL_CAPABILITIES, the source of truth for
isReasoningModel() on OpenAI models. Without an entry, reasoning defaulted
to false, so the UI never pinned temperature to 1 and the API rejected
temperature: 0 in both the prompt playground and online scoring.Add the three models with their none/low/medium/high/xhigh/max reasoning
effort options, adding "max" to the ReasoningEffort union and its label map.Co-authored-by: Claude Opus 4.8 noreply@anthropic.com
下载附件