Unverified50% confidenceFactExact time
Artificial Analysis consumed 160 million output tokens running the Qwen3.8-27B index, while the median for comparable open-weight models is 43 million, indicating that this model is verbose in its default reasoning.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/15/2026
First Seen
Valid until: 1/13/2027
Sources
Related Entities
Related Claims
UnverifiedArtificial Analysis noted that Qwen3.8 generated at least 7 million tokens during testing, while the median was only 820,000, indicating verbose reasoning.78% similarUnverifiedThe model places n-gram lookup tables at layers 1 and 14, with 24 hashes each indexing approximately 16 million rows, totaling 196 billion parameters, exceeding one-third of the model's weights.71% similarUnverified阿里千问新一代模型参数达2.4万亿，是该系列首个突破万亿的多模态大模型，代码生成准确率据称达到94.7%，较上一代提升约11个百分点71% similarUnverifiedThe full Neural 3 model with 121 million parameters has an actual computational load roughly equivalent to a 50-million-parameter model.71% similarVerifiedBefore outputting a final answer, a reasoning model's internal chain of thought may consume 3,000 to 20,000 tokens, far exceeding the approximately 500 tokens of an ordinary model.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1033149API
curl https://kongchang.com/api/v1/knowledge/claims/1033149MCP
get_claim(id=1033149)