o1-mini o1-preview Models ​

When using o1-mini or o1-preview, please note the following:

• Does not support system messages, streaming, image recognition, or the max_tokens field
• Replace max_tokens with max_completion_tokens
• The reasoning process consumes tokens; the actual output may not be large, but the combined reasoning tokens can exhaust your token budget
• Note: the maximum output for max_completion_tokens is 65536 for o1-mini and 32768 for o1-preview
• If max_completion_tokens is set too small (e.g., 4096), you will receive an empty response, but tokens will still be consumed; if set to the maximum (65536), a large number of tokens will be consumed even though the actual output may be small