3 3 hours ago

Qwen3.6-35B-A3B Uncensored is a Mixture-of-Experts model with 35.5B total parameters and 3B activated, extended to 1M-token context with the official MTP speculative-decoding layer, delivering 1.5x decode speedup on Ollama, with native thinking and tools.

tools thinking
6bbd26bdc087 · 94B
{
"num_ctx": 1010000,
"stop": [
"<|im_end|>"
],
"temperature": 0.6,
"top_k": 20,
"top_p": 0.95
}