AI Briefing#qwen
The latest AI and developer news about #qwen, with the original source and a short summary.
The latest AI and developer news about #qwen, with the original source and a short summary.
An uncensored Aggressive version of Qwen3.6-27B and K_P quantizations have been released.
Qwen has released Qwen3.6-27B, an open-source model in the 27B class.
Qwen 3.6 27B, a new model in Alibaba's Qwen series, has been released.
The little-coder and Qwen3.6-35B-A3B combo scored 78.7% on Aider Polyglot.
A Metal kernel implementation combining INT3 model compression with INT2 KV cache has been released.
In a 108K-token JS file memory test, Qwen 3.6 significantly outperformed Gemma 4.
Qwen has open-sourced Qwen3.6-27B, a 27B dense multimodal model.
Even with 32GB VRAM, llama.cpp's `--fit` ran Qwen3.6 Q8 at a 256k context at 57 t/s.
Running the same prompt across 9 local models revealed a correlation between quantization provider differences and code length.
Shared a setup for running Qwen3.6-35B-A3B with llama-server on an RTX 4060 8GB VRAM, along with a thinking budget pitfall.
Even 'uncensored' models have a flinch that lowers the probability of banned words on their own.
Qwen3.6-Max-Preview significantly boosts coding, knowledge, and instruction-following performance.
The DeepSeek-R1 family was compressed and fine-tuned for medical use with 4-bit GPTQ+QLoRA.
On an RTX 3090, Qwen3.5-27B with DFlash+DDTree recorded up to 207 tok/s.
Qwen 3.6 Max Preview has been posted on Qwen Chat, scoring 52 on the AA Index.
China's open-source AI models, including Alibaba's Qwen, are recording high market share worldwide.
Compressed the KV cache on Qwen 3.6's 1M context, reducing it from 10.7GB to 6.9GB.
Llama.cpp token speed measurements on the Zenbook A16 with Snapdragon X2 Elite Extreme.
Ran Qwen3.6 35B 4bit for multi-user use via vLLM Docker on 2x RTX 3090.
Qwen 3.6 35B-A3B has been quantized for Ollama, supporting even tools and thinking.
On DGX Spark, Qwen3.6 performs similarly to Qwen3.5, but MTP significantly increases throughput.