
Qwen3.6 Community Fine-Tune Hits 287K Downloads With Zero Refusals
HauhauCS ships a refusal-free Qwen3.6-27B build with custom K_P quants, native multimodality, and a 262K context tuned for long agent loops.
20 stories

HauhauCS ships a refusal-free Qwen3.6-27B build with custom K_P quants, native multimodality, and a 262K context tuned for long agent loops.

Poolside opens up with a 33B mixture-of-experts coder that activates only 3B parameters, runs on one GPU, and ships under Apache 2.0.

Deta open-sourced Surf, a local-first AI notebook that pulls PDFs, YouTube, tweets and files into a single research canvas with your choice of models.

A developer inverted OpenAI's Privacy Filter to extract structured PII spans from any text pile, using the same 1.5B weights and taxonomy.

DeepSeek open-sourced a library of GPU kernels written in TileLang that power its internal training and inference pipelines.

Z Lab's DFlash uses a diffusion model as a speculative drafter for Qwen3-27B, hitting over 6x lossless speedup and outpacing EAGLE-3 by 2.5x

Unsloth's Dynamic NVFP4 quant runs Qwen3.6-35B-A3B up to 1.79x faster on 32GB Blackwell GPUs while matching BF16 accuracy on MMLU-Pro, GPQA, and AIME.

Fudan University's FVL Lab released OpenOCR, an Apache-licensed toolkit bundling a 0.1B document parser, unified text/formula recognizer, and 24 reproduced STR methods.

OpenContracts turns a folder of documents into a programmable citation graph, exposing the same substrate to humans, apps, and MCP-compatible agents.

A 6.8K-star bilingual field guide to RL infrastructure, SGLang internals, and the messy engineering reality behind modern LLM training.

Ironclad's open-source Rivet gives you a node-based canvas for building, debugging, and shipping LLM agents as embeddable TypeScript graphs.

Microsoft's Playwright MCP gives any LLM a real Chromium browser through structured accessibility trees, not screenshots. It now powers 30+ AI clients.

Collaborator packages terminals, files, and running agents into a single infinite pan-and-zoom canvas, replacing the tab-switching dance of multi-agent coding.

OpenAI open-sources a 1.5B-parameter PII detection model that runs locally, strips sensitive data in one pass, and hits 96% F1 out of the box

A community INT4 quantization shrinks Qwen3.6-27B from 54GB to 18GB while keeping its Multi-Token Prediction head alive for 2x faster decoding.

A community AWQ INT4 quantization of Qwen3.6-27B shrinks Alibaba's flagship-tier coding model to about 20GB, making it self-hostable on a single consumer GPU.

An uncensored Qwen3.6-27B build hits 0/465 refusals with custom K_P quants, native vision, and a 262K context window in GGUF format.

A community fork of Claude Code drops the single-vendor lock-in, wiring the same terminal agent loop to OpenAI, Gemini, DeepSeek, Ollama and 200+ models.

A high-severity flaw in Flowise lets anyone on the internet mint OAuth access tokens from public chatflows without authentication. Patch is out.

Alibaba's new 27B dense open-weight model beats its own 397B MoE flagship on agentic coding, with Thinking Preservation and hybrid linear attention.