Tracked topic
Qwen3 is an AI model family developed by Alibaba, released as a set of large language models for natural-language tasks.
… The website relies on Qwen 3.5 models for benchmarking: Qwen3.5-9B – Required Small device baseline Qwen3.5-27B – Required Mid-range device baseline Qwen3.5-35B-A3B MoE – Optional MoE performance reference Qwen3.5-122B-A10B MoE – Optional Large memory device reference Qwen3.5-397B-A17B MoE – Option… …
… A fairly long list of models is provided: Vision – YOLO11, Yolo-World-V2, Yolov7-face, Depth-Anything-V2, MixFormer-V2, Real-ESRGAN, SuperResolution, RIFE Large Language – Qwen3-0.6B, Qwen3-1.7B, Qwen2.5-0.5B-Instruct, Qwen2.5-1.5B-Instruct, DeepSeek-R1-Distill-Qwen-1.5B, MiniCPM4-0.5B Multimodal –… …
… Qwen3-VL-2B benchmark and efficient data were also provided. …
… The company notably highlights support for Qwen3-256B-A22B and DeepSeek V3-671B. …
… So I restarted to give access to compute on the LAN instead, using 0.0.0.0 for the host: 1 jaufranc @ CNXSOFT - spacemitk3picoitx : ~ / LLM $ llama - server - m Qwen3 - 8B - Q4 K M . gguf - t 8 -- host 0.0.0.0 -- port 8080 Performance is about the same at 4.7 tokens per second; it just consumes few… …