Released on Sep 19, 2024
Near-70B quality at a fraction of the size: the sweet spot for advanced use. Excellent in Italian, code and reasoning. Needs 24GB+ VRAM.
Flagship Max da 2,4 trilioni di parametri con ~95B attivi, contesto da 1M token e primo modello Max aperto dei pesi, oltre al Qwen3.8-27B Apache 2.0.
Flagship Qwen3.5-397B-A17B basata su MoE sparsa con attenzione lineare (Gated DeltaNet), multimodalità nativa e 201 lingue.
Generazione successiva con modalità ibride di ragionamento (thinking e non-thinking), pretraining raddoppiato a ~36 trilioni di token e 8 taglie open-weight.
Rilasciata insieme alla famiglia Qwen2.5 con taglie da 0,5B a 72B, contesto fino a 128K token e licenza Apache 2.0.
Visit the official site to download, learn more and start using Qwen2.5 32B.
7B open-weight model (32-layer DiT) that unifies text-to-image generation up to 2048x2048, editing with up to 10 reference images, and native RGBA transparency. Designed to run locally on 16GB+ VRAM; supported by diffusers and ComfyUI.
Ultra-compact 0.5B-parameter model. Runs on any hardware, even embedded: ideal for lightweight summarization and low-power local devices.
Clear step up in reasoning and instruction quality over 8B models. Excellent for coding, agents and structured writing on 16GB GPUs or Macs with 32GB RAM.
