By skyforbes I’m currently trying to figure out how to manually adjust all the standard generation parameters that are mentioned in OpenAI’s documentation (and in academic papers I’m replicating). Specifically:
• temperature – randomness / creativity
• top_p – nucleus sampling filter
• top_k – number of candidate tokens considered per step
• presence_penalty – encourages new topics
• frequency_penalty – discourages repeating words
• max_tokens – controls output length
• stop / stop_sequence – defines where generation should stop
• seed – for reproducible results
The issue: since the new OpenAI Platform UI (mid-2025) was released, the classic Playground (with sliders for these values) seems to be gone.
Now I only see the Prompt Builder interface (with Developer Message, Prompt Message, Reasoning Effort, etc.), but there are no visible fields for temperature, top-p, penalties, or any other generation settings.
I understand that you can technically do this via the API or SDK (Python, JavaScript, .NET, etc.), but:
1. Is there any built-in UI left (or hidden chat mode) that still allows these parameters to be changed without coding?
2. If not, what’s the easiest workflow right now to send API requests with all parameters (especially top_k, presence_penalty, frequency_penalty, and seed)?
3. Does anyone know if OpenAI plans to reintroduce a Playground-like interface for testing prompts and parameter tuning?
For context:
• I’m using GPT-5-Nano / GPT-5 via platform.openai.com
• I’ve already created an API key and can access the Prompt Builder, but there’s no obvious way to tweak generation parameters
• I’m replicating results from a research paper where tuning these parameters (especially temperature and top_k) was required
Any insights, current workflows, or workarounds from people who’ve figured this out would be super helpful 🙏
Thanks in advance!