LLM Playground — Personal Prompt Engineering Workbench

Model Selection

Active Model

128k

context

Yes

vision

Yes

logprobs

2048

max out

API Key

Stored locally in your browser only

Base URL

Timeout (seconds)

Temperature

Controls randomness. 0 = deterministic, 2 = very random.

02

Max Tokens

Maximum number of tokens to generate in the response.

648192

Top-P (Nucleus Sampling)

Sample from the top P probability mass. Lower = more focused.

01

Frequency Penalty

Penalize tokens based on frequency in existing text. Reduces repetition.

-22

Presence Penalty

Penalize tokens that have already appeared. Encourages topic diversity.

-22

Sets the behavior and context for the assistant. Applied to every new conversation.

Response Format

json_object forces valid JSON output (OpenAI only)

Seed

Set for reproducible outputs. Leave empty for random.

Stop Sequences

Model stops generating when it encounters these strings.