Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
Orthrus injects trainable diffusion attention into frozen Qwen3 backbone, achieving 7.8× token/forward speedup with identical output distribution.
User reports local agent (Pi + Qwen 3.627B) executed rm -rf on full disk, demonstrating autonomous decision-making risk in unmonitored coding agents.
Qwen 3.6 35B multi-token prediction variant achieves ~1.5x speedup over baseline in 300k-token context local inference experiments with pygame project.
RTX 5060 Ti local LLM configuration guide covering vLLM and llama.cpp serving of Qwen 27B/35B models with quantization and long-context presets.
User documents quantization recipe for Qwen3.6 27B INT8 model showing improved code generation despite reduced reasoning tokens.
Researcher implements automated red-teaming loop using RL to generate diverse jailbreaks against Qwen3.5, then uses discovered attacks to improve model defenses.
User asks about quantization quality differences (Q4 vs Q6) for Qwen 3.6 on consumer GPUs; anecdotal discussion, no new data.
Hugging Face releases ml-intern, an agent framework for local LLM research automation supporting llama.cpp/ollama with Qwen and Claude models.
Shipped this for the AMD x lablab hackathon. Attached video is one of the actual reels the pipeline produced - one English sentence in, finished mp4 with characters, story, music, and voice-over out (fast demo video, not the best quality). ~45 minutes end-to-end on a single AMD Instinct MI300X. Every model is Apache 2.0 or MIT. **Pipeline (8 stages, all sequential on the same GPU):** 1. **Director Agent** - Qwen3.5-35B-A3B (vLLM + AITER MoE) plans 6 shots from one sentence, returns structured JSON with character bibles, shot prompts, music brief, per-shot voice-over script, narration langua...
Multi-token prediction + TurboQuant quantization achieves 40% throughput gain (21→34 tokens/s) on Qwen 27B/35B via LLaMA.cpp on M-series Mac.
Qwen 3.6 35B and Gemma 4 26B MoE models achieve 20–24.5 tok/s on GTX 1080 with 128k context via llama.cpp quantization.
Qwen 3.6 27B achieves 52.8 tokens/sec throughput on MI50 GPU with full precision, enabling on-device agent inference.
User demonstrates practical applications of local LLMs including embedding-based semantic search and Qwen 35B for automated database evaluation workflows.
Reddit discussion comparing Qwen 3.6 27B vs. 100B+ open models for local coding; user survey of high-VRAM model preferences.
Luce ships DFlash+PFlash optimizations for AMD Ryzen AI MAX+ 395, achieving 2.23x decode speedup on Qwen 3.6-27B vs llama.cpp HIP.
Developer demonstrates local coding workflow using Qwen2.5-Coder-7B autocomplete and Qwen3.6-35B agentic model on RTX 5080 with RAM offloading.
Empirical study of supervised fine-tuning on procedural skills across Qwen3.5 0.8B-4B models shows W-shaped pre-training trajectory and uniform SFT gains.
Reddit user benchmarks open-weight LLMs (Qwen 3.6, Zaya1) on chess visualization task; Qwen 35B-A3B achieves near-perfect SVG chessboard generation.
Qwen-Scope: open-source suite of sparse autoencoders for mechanistic interpretability across 7 Qwen models.
User shares tap-water cooling setup for DGX running Qwen3.5-122b, achieving 68°C at 95% utilization with 18.77 tokens/sec.
Empirical study across 288 model calls identifying JSON output failures in Llama 3, Mistral, Command R, DeepSeek, Qwen; failure modes consistent across open and closed models but vary by rate.
Reddit speculation on whether Alibaba will release larger Qwen3.6 models or specialized coding variants.
Qwen3.5 0.8B sees 2.88M monthly downloads; user reports semantic understanding, JSON parsing, and latency challenges in production workflows.
User reports Qwen 3.6 35B outperforms Gemma 4 26B on speed and long-context, with better prompt adherence in local inference.
Unsloth released GGUF quantizations of Qwen3.6 27B and 35B models with MTP training.
JSON parsing bug in llama-server's chat-template-kwargs for Qwen3.6 preserve_thinking parameter requires whitespace-free formatting.
Reddit user reports Qwen 3.6 35B outperforms prior small models (Devstral) on niche academic code understanding tasks.
User documents running Qwen3.6 35B A3B quantized model on RTX 4060 (8GB VRAM) with 190k context window at 37-51 tok/sec throughput.
User describes using Qwen 35B locally on a flight to troubleshoot Linux DNS/WiFi connectivity issue via agent interaction.
Hugging Face co-founder claims Qwen 3.6 27B matches Claude Opus on coding tasks offline, signaling convergence in local model capability.