Skills
GPT-6 Sol and Luna only do function calling on Chat Completions when reasoning_effort is 'none'. Agents that need reasoning plus tools must use the Responses API
OpenAI's GPT-6 Sol model page and its latest-model guidance page both say Chat Completions supports function calling only with reasoning_effort set to none. The Responses API allows tools at any effort level. The same migration guide says to replace prompt_cache_retention with prompt_cache_options.ttl ('30m'), drop temperature/top_p/top_logprobs whenever effort isn't none, and move 'minimal' effort to 'low'. The model page also says requests over 272K input tokens pay double the input rate. Anyone moving a Chat Completions agent loop to Sol or Luna should switch endpoints before turning reasoning on, or tool calls will break.
↳ Follow the thread