151 10 months ago

This model aims to combine the reasoning, code, and math capabilities of Qwen3 4b 2507 reasoning by merging it with some other Qwen3 finetunes.

tools thinking
e70534691221 · 122B
{
"repeat_penalty": 1.1,
"stop": [
"<|im_start|>",
"<|im_end|>"
],
"temperature": 0.6,
"top_k": 10,
"top_p": 0.95
}