China 8.2B parameters Text generation

Explore Qwen3-8B

Write a prompt, choose how the model reasons, and fine-tune every useful generation control. Advanced settings start with Qwen's recommended thinking-mode defaults.

Model

Qwen/Qwen3-8B

Conversation

What should Qwen work on?

System prompt Optional

Set the model's role, tone, or rules for this response.

Your message

Required0 / 32,768 characters

Parameters

Generation settings

Thinking modeLet Qwen reason before answering. Sample outputDisable for greedy decoding.

Temperature

Controls randomness.

Top P

Samples from cumulative probability.

Top K

Limits the candidate token set.

Min P

Filters very unlikely tokens.

Presence penalty

Reduces repeated topics and phrases.

Repetition penalty

Discourages repeated text.

Maximum output tokens

Between 1 and 32,768.

Seed Optional

Reuse a seed for more repeatable output.

Stop sequences Optional

Generation stops when any sequence appears.

Explore Qwen3-8B

What should Qwen work on?

Response