GPT-3.5-turbo-16k: Word limit

API
1

Hi guys,

I have an application that generates articles optimized for SEO using the API and I noticed that lately the model’s responses do not meet the user’s request regarding the number of words. Ex: the user asks for an article with more than 600 words and I receive around 300 at most. Are there any limits in this regard?

const response = await openai.chat.completions.create({
  model: "gpt-3.5-turbo-16k",
  messages: [
    {
      "role": "system",
      "content": "You are a wizard who generates SEO-optimized articles. \nMake sure to separate the paragraphs with the <p> tag and highlight the title whenever it appears. \nWhen asked for a number of words, deliver a larger number. \nFollow SEO best practices to provide rich, well-optimized content."
    },
    {
      "role": "user",
      "content": "Generate an SEO-optimized article using the title: \"AI revolution\"  as a reference.\nIdentify and highlight the keywords that are repeated in the article with the <strong> tag.\nParagraphs must be separated by the <p> tag.\nSome observations about the article: \"The article must contain more than 600 words.\""
    },
],
  temperature: 0.7,
  max_tokens: 8019,
  top_p: 1,
  frequency_penalty: 0,
  presence_penalty: 0,
});