Qwen 3.5
Qwen 3.5 is a large language model in Alibaba Cloud's Qwen series, representing an iterative version within that family. The model supports natural language understanding and generation, multi-turn dialogue, logical reasoning, and code generation, with multilingual interaction capability. It is applicable to a variety of natural language processing tasks including text creation, question answering, and information extraction, and can be deployed via API or as an open-source release.
Core Facts
Qwen3.6-27B supports multimodal capabilities, including text reasoning, image understanding, and video processing.
Qwen3-series coding models load at approximately 20 GB and can expand to 35–40 GB during inference.
Qwen (Tongyi Qianwen) is a large language model series developed by Alibaba DAMO Academy; it has been continuously open-sourced and iterated since 2023 under the Apache 2.0 license.
Timeline (last 90 days)
Ash's LLM Picker project collected 92 public Mac measurement data points, showing that on most models MLX leads LLAMA.cpp by only 1 to 1.2 times, with the large gaps mainly concentrated in the Qwen 3.5 and 3.6 series.
PewDiePie tried fine-tuning a small-parameter model from the Qwen 3.5 series to work with Odysseus locally for tool calling.
In actual testing, the 9B version of Qwen 3.5 was unable to complete the task of creating a sorting algorithm visualization webpage, spending over half an hour repeatedly rewriting code that ultimately failed to run.
The 9B version of Qwen 3.5 succeeded at the small task of adding a copy button to a password manager input field, correctly implementing the copy-to-clipboard functionality and autonomously generating an SVG icon.
Small 9B models struggle with complex tasks requiring planning and extensive code generation, but are well-suited to small tasks with clearly defined scope.
The 397B version of Qwen 3.5 run via Ollama's cloud successfully completed the sorting algorithm visualization task, generating approximately 900 lines of code with the page running correctly without errors.
Kev is an open-source classification model based on Qwen 3.5's 0.8B base model, with LoRA fine-tuning weights merged in, using unidirectional reading and making decisions through a pointer head.
Kev is an open-source reproduction project built on Qwen 3.5, offering decision models in three sizes: 0.8B, 4B, and 9B.
Baseline tests shown by the UP host indicate that MiniCPM-2B outperforms 4B models such as the 4B version of Qwen 3.5 and Gemma 4B in code reasoning and mathematical reasoning dimensions (the verbal presentation contains ambiguous wording; refer to official data for accuracy).
Qwen (Tongyi Qianwen) is a large language model series developed by Alibaba DAMO Academy; it has been continuously open-sourced and iterated since 2023 under the Apache 2.0 license.
2 more timeline events
All Facts (13)
Qwen3.6-27B supports multimodal capabilities, including text reasoning, image understanding, and video processing.
80%VerifiedQwen3-series coding models load at approximately 20 GB and can expand to 35–40 GB during inference.
70%VerifiedQwen (Tongyi Qianwen) is a large language model series developed by Alibaba DAMO Academy; it has been continuously open-sourced and iterated since 2023 under the Apache 2.0 license.
65%VerifiedQwen3-4B is already on par with DeepSeek V3 or GPT-4o on multiple reasoning evaluation metrics.
65%UnverifiedAsh's LLM Picker project collected 92 public Mac measurement data points, showing that on most models MLX leads LLAMA.cpp by only 1 to 1.2 times, with the large gaps mainly concentrated in the Qwen 3.5 and 3.6 series.
50%UnverifiedThe 397B version of Qwen 3.5 run via Ollama's cloud successfully completed the sorting algorithm visualization task, generating approximately 900 lines of code with the page running correctly without errors.
50%UnverifiedIn actual testing, the 9B version of Qwen 3.5 was unable to complete the task of creating a sorting algorithm visualization webpage, spending over half an hour repeatedly rewriting code that ultimately failed to run.
50%UnverifiedThe 9B version of Qwen 3.5 succeeded at the small task of adding a copy button to a password manager input field, correctly implementing the copy-to-clipboard functionality and autonomously generating an SVG icon.
50%UnverifiedSmall 9B models struggle with complex tasks requiring planning and extensive code generation, but are well-suited to small tasks with clearly defined scope.
50%UnverifiedKev is an open-source classification model based on Qwen 3.5's 0.8B base model, with LoRA fine-tuning weights merged in, using unidirectional reading and making decisions through a pointer head.
50%UnverifiedBaseline tests shown by the UP host indicate that MiniCPM-2B outperforms 4B models such as the 4B version of Qwen 3.5 and Gemma 4B in code reasoning and mathematical reasoning dimensions (the verbal presentation contains ambiguous wording; refer to official data for accuracy).
50%UnverifiedPewDiePie tried fine-tuning a small-parameter model from the Qwen 3.5 series to work with Odysseus locally for tool calling.
50%UnverifiedKev is an open-source reproduction project built on Qwen 3.5, offering decision models in three sizes: 0.8B, 4B, and 9B.
50%