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.
By applying MTP to the Qwen3.6-35B-A3B model, a decoding speed improvement of about 1.49x was achieved on llama-server.
Qwen has released Q-Judger, a VLM model that automatically evaluates text-to-image generation quality.
The ReAligned-Qwen3.5 model series, tuned to mitigate Chinese ideological bias and censorship, has been released.
Alibaba Cloud declared the era of agentic AI, unveiling the new model Qwen 3.7 Max and agent-dedicated cloud infrastructure.
It has been mathematically proven that various AI robustness techniques all estimate the same 'deployment nuisance covariance matrix.'
hipEngine, an open-source engine that maximizes inference performance for Qwen 3.6 models on AMD RDNA3 GPU environments, has been released.
A bug in llama.cpp causes the KV cache of the Qwen3.6-35B-A3B model to be dropped, significantly degrading inference speed.
Native MTP support for Qwen3.6 MLX models has been added to the distributed inference tool exo.
The BeeLlama v0.2.0 update improves inference speed for Qwen 3.6 and Gemma 4 models by up to 4.9x.
The BeeLlama v0.2.0 update significantly boosts inference speed for Qwen and Gemma models through DFlash optimization.
Numind has released NuExtract3, a VLM specialized in document structuring and OCR based on Qwen3.5-4B.
Qwen3.7 Max's intelligence index rose compared to the previous model, showing major advances especially in scientific reasoning and coding.
Qwen3.7-Max showed top-tier performance on coding, reasoning, and agent benchmarks.
Alibaba unveiled its new Zhenwu M890 chip for AI agent workloads and announced a chip roadmap through 2028.
News has emerged that the Qwen team is likely to release a new model at the 27B scale.
ByteShape released Qwen 3.6 35B GGUF NTP and MTP quantizations along with benchmarks.
Qwen3.7-Max has unveiled its agentic, coding, and long-horizon autonomous execution performance.
The sharp decline in AI inference costs is being driven by software, and local open models are quickly catching up to frontier model performance.
Qwen has unveiled Qwen3.5-LiveTranslate, a real-time interpretation model that leverages visual context as well.
Qwen3.5-9B's political censorship was found to be a small circuit that decomposes into 3 axes.
Qwen3.7 Preview launched on Arena's Text and Vision categories, ranking 13th and 16th respectively.
MTP speculative decoding has been added to llama.cpp, making Qwen3.6 27B up to 2.44x faster.
TIME methodology and TIMEBench have been released, training Qwen models to perform short reasoning only when needed.