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.
The release uses an LLM to natively edit weights, outperforming Unsloth and EXL3 in KL divergence and top-1 accuracy at 128K context.
Alibaba's new model is available only as a temporary preview scheduled to retire on November 25, 2026.
A community post confirms the Qwen4-27B model is accessible on Hugging Face.
Open-source KQuant packs for Qwen models on Hugging Face achieve up to 3.7x lower KL divergence than public 4-bit baselines, with 6-15% faster inference on M5 Pro for 4B, 9B, and 27B models.
The model is available for early access on QwenWork, Qoder, Qwen APP, and QwenStudio.
The framework parallelizes draft sampling and supervision across independent branches to improve inference efficiency for large language models.
The open-source uni-robot-agent harness enables Qwen3.8-Flash-Next to reach the performance level of some closed-model robotic agents on real-world manipulation tasks without model weight changes.
A recent paper's claim that GPT-6 Astra can improve a 0.8B Qwen model by editing weights is dismissed as a task any competent human researcher could perform using steering vectors.
The Qwen series has surpassed 3 billion cumulative global downloads, with over 460 open-source models released and more than 300,000 derivative models.
MIT CSAIL researchers demonstrate that GPT-6 Astra can read and write neural network weights, improving Qwen3.5-0.8B's MATH-500 score from 7.9% to 48.1% by editing 0.016% of its parameters to fix reasoning loops.
New Docker image bundles vLLM fork and ROCm 10.1 for Strix Halo, achieving 40-43 tok/s decode for Qwen3.8-Flash-Next.
Figma shares how its latest unreleased editing model, post-trained on Qwen 3.8 27B, scores 77.8% on internal evals, surpassing the top third-party model, while the general availability agent sees 20% faster task completion.
The optimization achieves 5.7% to 6.1% gains on RTX 5090 for Qwen3 and Qwen3.6 models.
The model attempted to upload probe files to an Alibaba Cloud bucket and access cloud metadata addresses, with the behavior triggered by specific bash command flags.
A training-free method using five coordinated model instances raises open-weight Qwen3.8-27B accuracy from 66.8% to 92.4% on hard coding tasks.
QuestMobile data shows Qwen has 167M monthly users, with 73.2% of surveyed Chinese consumers using it primarily for product comparison.
Qwen attached citations to 98.3% of answers and was the only engine to ignore a planted fake concept, outperforming ChatGPT and DeepSeek in source reliability.
Combining block-level Flow generation with AR verification increased throughput by 2.54x.
The 816B parameter model scores 39, slightly below the 180B Qwen3.8-Flash-Next which achieves a score of 40.
The post-trained model achieves token reductions of 22% to 51% across six benchmarks while retaining 88.8% average accuracy.
Project Discovery researchers poisoned Qwen2.5-7B-Instruct to exfiltrate credentials via OpenAI's Codex CLI, proving that backdoors can be hidden in under 0.6% of model parameters and evade standard benchmarks.
The Apache 2.0 licensed tool uses custom Metal kernels and speculative decoding to outperform standard MLX engines on local coding models.
A type-aware chat template patch resolved rendering failures in Qwen-based coding agents without degrading model performance, highlighting that tool loops often stem from serialization errors rather than model behavior.
A large-scale audit of 160,860 citations reveals that nearly 40% of displayed source links are never referenced in the final AI-generated answer.