Qwen3.5 27B
Open weightsMultimodalReasoningTool useVision
Qwen3.5 27B is an open-weights vision-language model from Alibaba's Qwen team that accepts text, image, and video input. It targets visual reasoning, document understanding, and agent tasks, with a 262k-token context window.
256Kcontext
$0.3 / $2.4per 1M tokens (in / out)
27.8Bparameters
5stories in our feed
Specs
hover a value to see its sourceContext window256K
Max output64K
Parameters27.8B
Knowledge cutoff—
Licenceapache-2.0
Commercial useCommercial OK
InputText · Image · Video
OutputText
HF downloads (30d)1.8M
Familyqwen
API price
USD per 1M tokensInput$0.3
Output$2.4
Cached input—
No price change recorded since we started tracking.
News timeline
from the AITOP feed, one story per event2026-10-28
形式化器打通开放任务与快速决策模型：CALVIN-D 成功率提升至 52.6%arXiv: DeepSeek
2026-10-09
StoreBench Live-Commerce Agent EvaluationarXiv: DeepSeek
2026-09-29
Meta RAM 团队提出 RL-XAR：用专家对齐评分标准做写作 RLshao__meng
2026-09-04
2026-08-05
SWD：稀疏权重分解让电路提取更高效arXiv cs.LG