What are the odds Qwen 4 models are going to think less?
Summary
该推文猜测即将发布的 Qwen 4 系列模型是否会减少思考（thinking）token 的消耗，作者认为在保持或提升智能的同时少思考是最大的升级方向。
Similar Articles
Unpopular opinion : Qwen 3.8 27b is not an overthinker
The article argues that Qwen 3.8 27b's increased reasoning token usage is similar to other Chinese AI models like GLM and DeepSeek, with user frustration stemming from hardware limitations. It suggests using a reasoning budget can maintain performance over Qwen 3.6.
LessThink-Qwen3-4B: the same model, with far less thinking [P]
The author post-trained Qwen3-4B to reduce token usage by 44% in reasoning while preserving knowledge and style, with training completed on a single GPU.
New model from Telperion: Qwen3.8-27B-MLA & Qwen3.8-27B-GLA-g2. 4x smaller KV cache, smarter, and more efficient reasoning.
Telperion 发布了 Qwen3.8-27B 的潜在注意力（MLA 与 GLA-g2）变体，将 KV cache 压缩 4 倍，在消费级硬件上实现接近 4 倍的最大上下文或服务器端 4 倍并发，并在 SWE-bench Verified 上取得 +6.58% 的性能提升，同时减少约 20% 的推理 token 消耗。
@_akhaliq: bottlecapai/ThinkingCap-Qwen3.6-27B Capability of Qwen3.6-27B with 50% less thinking tokens on average, and over 90% le…
bottlecapai releases ThinkingCap-Qwen3.6-27B, a finetuned version of Qwen3.6-27B that achieves 50% fewer thinking tokens on average and over 90% fewer in best cases, improving efficiency.
Qwen3.8-27B different thinking levels
The Qwen3.8-27B model is introduced with varying thinking levels, showing improved reasoning capabilities compared to previous versions like Qwen 3.7 plus and Qwen3.6-27B.