albedo-qwen3.6-35b

A Hugging Face–hosted derivative or repackaging of Alibaba's Qwen3.6-35B-A3B open-weight model, published under the power612 account. The underlying model is a 35B-parameter mixture-of-experts architecture with only 3B parameters active per token, supporting multimodal (text + vision) input, long-context reasoning up to ~262K tokens, and agentic tool-use workflows. Source: supabase.