{"id":654,"date":"2025-04-09T01:43:09","date_gmt":"2025-04-09T01:43:09","guid":{"rendered":"https:\/\/www.juhenext.com\/?post_type=model&#038;p=654"},"modified":"2025-04-09T01:44:01","modified_gmt":"2025-04-09T01:44:01","slug":"qwq-32b","status":"publish","type":"model","link":"https:\/\/www.juhenext.com\/zh\/model\/qwq-32b\/","title":{"rendered":"QwQ-32B"},"content":{"rendered":"<p>QwQ-32B is a medium-sized reasoning model from the Qwen series, optimized for enhanced performance in downstream tasks, particularly challenging problems requiring deep reasoning. Unlike conventional instruction-tuned models, QwQ-32B integrates advanced architectural components such as RoPE, SwiGLU, RMSNorm, and Attention QKV bias. With 64 layers, 40 query heads, and 8 key-value heads (GQA), it supports a full 131,072-token context length, though YaRN must be enabled for prompts exceeding 8,192 tokens. Pretrained and post-trained via supervised finetuning and reinforcement learning, it achieves competitive results against leading models like DeepSeek-R1 and o1-mini. Users can explore its capabilities via QwenChat or refer to official resources for deployment guidelines.<\/p>","protected":false},"excerpt":{"rendered":"<p>QwQ-32B \u662f Qwen \u7cfb\u5217\u4e2d\u7684\u4e00\u4e2a 32.5B \u53c2\u6570\u63a8\u7406\u6a21\u578b\uff0c\u5177\u6709\u5148\u8fdb\u7684\u67b6\u6784\u548c 131K \u4ee4\u724c\u4e0a\u4e0b\u6587\u957f\u5ea6\uff0c\u65e8\u5728\u5728\u590d\u6742\u4efb\u52a1\u4e2d\u8d85\u8d8a\u50cf DeepSeek-R1 \u8fd9\u6837\u7684\u6700\u5148\u8fdb\u6a21\u578b\u3002<\/p>","protected":false},"featured_media":549,"template":"","meta":{"_acf_changed":false},"context-window":[64],"features":[16,24,15,23,19],"maximum-output":[31],"model-type":[11,29],"promotion":[],"provider":[58],"recommend":[63],"class_list":["post-654","model","type-model","status-publish","has-post-thumbnail","hentry","context-window-32k","features-function-calling","features-reasoning","features-streaming","features-text-input","features-text-output","maximum-output-8k","model-type-chat","model-type-reasoning","provider-qwen","recommend-budget-friendly"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/model\/654","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/model"}],"about":[{"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/types\/model"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/media\/549"}],"wp:attachment":[{"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/media?parent=654"}],"wp:term":[{"taxonomy":"context-window","embeddable":true,"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/context-window?post=654"},{"taxonomy":"features","embeddable":true,"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/features?post=654"},{"taxonomy":"maximum-output","embeddable":true,"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/maximum-output?post=654"},{"taxonomy":"model-type","embeddable":true,"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/model-type?post=654"},{"taxonomy":"promotion","embeddable":true,"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/promotion?post=654"},{"taxonomy":"provider","embeddable":true,"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/provider?post=654"},{"taxonomy":"recommend","embeddable":true,"href":"https:\/\/www.juhenext.com\/zh\/wp-json\/wp\/v2\/recommend?post=654"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}