Tongyi Qianwen
29 articles tagged with this topic
Qwen 27B hits 50 tok/s: 16GB consumer GPUs can now run large models locally
A Reddit user ran Alibaba's Qwen 27B on a 16GB consumer GPU, hitting 50 tok/s generation + 100k context. Local LLMs are leaving the geek circle behind
Qwen3.8 Hits 94% on Mac — But Benchmarks Are Breaking Down
Alibaba quietly released Qwen3.8; a developer hit 94% on a ~$7,000 Mac. More telling: the blogger admits "models are getting too good to differentiate
User Claims Qwen Beats DeepSeek — But the Post Has Zero Content
A Reddit post claims Qwen3.8-Flash-Next beats DeepSeek V4 Pro-no benchmarks, no tables. We unpack the anxiety behind this empty post.
Qwen 27B Compressed to 3-bit Runs 3D Apps Locally — Open Source Closes Cloud Gap
Reddit LocalLLaMA compressed Alibaba's Qwen 27B with 3-bit IQ3XXS, running a 3D demo locally. Small-model + quantization + local trends accelerate.
Qwen 27B Fixed a Real Git Repo — But Its Eval Method Is the Real Story
A dev tested Qwen 27B on a real Git repo. It fixed code, ran tests, recovered. The real story: the eval method—and 6 criteria China AI lacks.
Qwen Autonomously Writes a C Compiler — Open-Source Agents Enter the Long-Haul Race
Qwen 3.6 27B built a C99 compiler autonomously in 6 weeks. Open-source handles long-haul tasks, but hallucinations and context issues still put "auton
No One Explains How to Run Qwen3 Locally—So a Reddit User Spent $100
A Reddit user is spending $100 to benchmark Qwen3 quantizations—exposing the open-source LLM ecosystem's failure to guide local-deployment users.
Qwen 27B Quantization Test: Only 0.2% Accuracy Loss on a Single GPU
A Qwen 27B compression test on a single RTX 6000 found Q6 accuracy just 0.2% below Q8 while saving ~4GB VRAM. The local-LLM barrier is falling fast.
Qwen Quantization Quality Varies 4×—Size Alone Isn’t Enough
A Reddit user tested 24 Qwen3.8-27B builds and found up to a 3–4× quality gap at 4-bit. File size alone is not a reliable guide.
Qwen 27B Lets Local AI Match Gemini — China's Small-Model Reversal
Qwen 27B matches or beats Gemini on OCR and code, per Reddit. A US team eyes self-hosted AI with under-two-month payback. Local AI feels usable.
Qwen 3.8-27B One Week In: Top Marks for Doing, Memory Slips
Qwen 3.8-27B earned 'local best' on agent tasks in 2,000 Reddit tests but regressed on knowledge memory—first open-source model approaching GPT on exe
Qwen 27B Beats OpenCode Locally — Framework May Limit Coding AI More Than Model
Reddit user tested Qwen 27B on RTX 3090: PI Agent beat OpenCode on quality, tokens, context. Coding AI bottleneck may be framework, not model.
Free Qwen Model Scores 52 — Halve Your ChatGPT Bill
Alibaba's new Qwen3.8 27B scored 52 on an independent AI benchmark — roughly matching early GPT-4, but completely free. For indie founders using AI da
Qwen 3.8 27B Benchmark Surges 37% — Open-Source Cracks Closed-Source Ceiling
Qwen 3.8 27B hits 52 on Artificial Analysis, up 37% from the prior version and surpassing Claude Opus 4.5 (42). The open-source vs. closed-source ceil
Qwen 27B vs GPT-5.6? Why a Single Reddit Post Went Viral
A Reddit post claiming "Qwen 27B = GPT-5.6 Luna compressed" went viral with no benchmarks. Why zero-evidence claims still explode—and what they signal
Qwen Thinks 90 Min, No Answer — One Parameter Cures Open-Source Overthinking
Alibaba's Qwen3.8-27b overthinks by default — inferences can run 90 minutes without output. A Reddit dev shared two llamacpp params that fix it.
Qwen Community Buzzes Over 9B — Alibaba's Open-Source Pace Leaves Rivals Behind
A Reddit 'Qwen 3.8 9b?' post caught our eye. With 10+ Qwen variants in 12 months, speculation reveals Alibaba's real open-source influence.
Alibaba's Qwen3 27B Compressed to 18GB, Runs on Single GPU — Local LLM Bar Drops Again
Alibaba's Qwen3 27B compressed to 18GB via int4 quantization with MTP acceleration, runs on a single consumer GPU. Local LLM hardware costs keep falli
Qwen 27B Learns to Fix Its Own Code — Open Source Crosses the Agent Deployment Threshold
Alibaba's Tongyi Qianwen 27B model shows a dramatic jump in agent self-correction between two minor versions. Open-source small models are closing the
Alibaba's Qwen 27B: Three Iterations, A Visible One-Shot Lift
Qwen 27B across three versions: 2.46 → 3.00 on 35 one-shot tasks. Small gains, stable direction. Worth watching for open-source LLM watchers.
Alibaba Qwen Drops 27B Open-Source — On-Prem AI Is Finally Within Reach
Qwen 27B is now open-source, free for commercial use. The community built consumer-grade versions within 24 hours—27B hits the capability-cost sweet s
Alibaba Qwen Redefines Multimodal: The Agent-First Shift Has Begun
Alibaba Qwen's livestream took "Agent First" as its theme, redefining multimodal from understanding to doing. The first major Chinese player to align
Alibaba's Qwen 27B Pushes Private AI From Demo to Budget
Alibaba's Tongyi Qianwen releases a 27B-class model. A Reddit user asked for help benchmarking local performance — behind that ~100-word plea lies the
Qwen3.8-27B Drops Early — 27B Is Local AI's Sweet Spot, Benchmarks Pending
Alibaba's Qwen team posted the Qwen3.8-27B model card on Hugging Face ahead of benchmarks. 27B is open-source's local AI sweet spot.
Qwen 3.8 Launches with 5 Bugs—Community Devs Ship a Universal Fix in One Week
Alibaba's Qwen 3.8 launched with adjustable reasoning depth but shipped with 5 critical chat template bugs. Community dev froggeric released a univers
Qwen 3.5 Hits 18 token/s on a $280 Radeon 7600 — Local LLMs Are Finally "Good Enough"
Alibaba's Qwen 3.5 35B MoE model runs at 18 token/s on a ~$280 Radeon 7600 via llama.cpp, pushing local LLMs into genuinely usable territory.
Old Tesla V100 Runs New Qwen3-6 27B: Local AI Deployment Costs Drop Again
Alibaba's Qwen3-6 27B model runs smoothly on a second-hand Tesla V100, driving a coding Agent. The bar for running large models locally is being lower
Qwen3.6 35B Beats 27B in Speed and Quality: Parameter Count Is Unreliable
Developers found Qwen3.6 35B outperforms 27B in quality and speed, breaking the "smaller is faster" myth. Benchmark data, not parameter counts, should
Tongyi Qianwen Replicates Deep Research in 200 Lines: Agent Moats Are Shallow
LangChain + Tongyi Qianwen replicate OpenAI's Deep Research in 3 steps, showing Agent barriers are low—but the demo-to-product gap remains.