Tagged articles
Qwen
83 articles · Page 1 of 1
The article debunks a viral screenshot claiming 7 Qwen4 models as fan speculation, but confirms Qwen's dominance in local LLM deployment via Ollama download stats (10/30 top models), third-party benchmarks ranking Qwen3.8-27B best, and a comprehensive Apache 2.0 size gradient. It provides Mac-specific model recommendations based on memory (16GB to 128GB+) and warns about license variations.
Apache 2.0 licenseMac hardwareOllama
0 likes · 10 min read

The article analyzes the July 2026 open-weights letter signed by 25 organizations, clarifies the critical difference between open weights and open source, examines licensing traps, and benchmarks current model sizes against Mac unified memory — revealing that the most 'open' model (1.56 TB) needs 12+ GPUs while the practical ceiling for a 24 GB Mac is a 16.1 GB Qwen 27B quant.
AI licensingHuang JensenKimi K3.1 Pro
0 likes · 17 min read

Alibaba's Qoder platform quietly added an anonymous model called Qoduck, free for personal users for a limited time, speculated to be a Qwen 4 Max preview, offering strong cross-file refactoring, long-context understanding, and agent-like planning with no usage credits required during the trial.
AI coding assistantAlibabaLong Context
0 likes · 5 min read

Controlled experiments comparing MIT's Astra weight-editing approach with a hypernetwork method reveal that while Astra fixes repetition loops, it fails to internalize knowledge, personas, or skills, unlike hypernetworks which succeed; the authors argue weight editing alone is insufficient for stable learning and propose a complementary agent-hypernetwork workflow.
AstraContinual LearningHypernetworks
0 likes · 15 min read

A Japanese developer implemented Qwen Sparse Attention (QSA) on Qwen3.8-27B, achieving 2.92x faster prefill and 1.79x faster decode at 261K tokens while maintaining perplexity and DocVQA scores; a quantized version further reaches 2375 tok/s prefill and 115 tok/s decode with 22.8 GiB VRAM.
Community DevelopmentInference OptimizationLong Context
0 likes · 5 min read

ByteDance releases Axion 1.0, a 122B parameter model achieving 66.7 on FrontierScience Research benchmark, surpassing GPT-5.6 Sol's 63.3, using only two supervised fine-tuning stages and reinforcement learning on Qwen base models, with 35B weights open-sourced under Apache 2.0.
Axion 1.0ByteDanceFrontierScience Research
0 likes · 5 min read

MIT researchers demonstrate GPT-6 Astra directly modifying Qwen3.5-0.8B's weights, improving MATH-500 scores from 7.9% to 48.1% (6x) by altering only 0.016% of parameters without gradient descent, and even writing neural network weights from scratch for games, outperforming PPO baselines 20x.
GPT-6 AstraMIT researchQwen
0 likes · 9 min read

Apple partners with Alibaba to integrate Qwen into China's Apple Intelligence starting with iOS 27.3 in December, but the system relies on cloud inference, excludes Siri AI and Mac support; meanwhile, high-memory Macs can locally run open-source models up to 405B parameters using MLX and llama.cpp, offering privacy, offline use, and model freedom.
Apple IntelligenceChina AI MarketCloud vs Edge AI
0 likes · 7 min read

MIT CSAIL's Weight-Space Programming lets an LLM read and modify neural network weights directly, without gradients or training. On Qwen3.5-0.8B, modifying 0.016% of weights raised MATH-500 from 7.9% to 48.1% by removing reasoning loops, showing LLMs can debug and edit model parameters.
AI researchLLMMIT CSAIL
0 likes · 17 min read

Meituan and Peking University propose ClawLoop, a Harness-native framework closing the loop of task generation, model training, and diagnostic feedback, enabling agents to self-improve without stronger teacher models; experiments on Qwen3.6-27B show consistent gains across three benchmarks after three iterations, with SkillsBench rising from 46.3 to 57.2.
Agent Self-ImprovementClawLoopGRPO
0 likes · 11 min read

B.AI's weekly token usage rankings for Oct 16-22 show DeepSeek leading Xiaomi by only 165.4B tokens, MiniMax surging 5200% from a small base, Qwen closing within 0.18% of Zhipu for third place, while three major Flash models offer simultaneous discounts, signaling intense price and market share competition among Chinese AI providers.
AI model APIDeepSeekFlash models
0 likes · 13 min read

Analysis shows HiRoute's intelligent model routing saves over 90% costs despite cache loss when switching models, because switches are infrequent, one-time losses amortize over many requests, cheaper models often cost less even with full reprocessing, and cache discounts only affect input while savings come from output-heavy cheaper models.
AstraHiRouteKV Cache
0 likes · 11 min read

The author builds an open-source pipeline using three skills — z-web-pack, z-video-downloader, and z-video-study-webpage-qwen — to download videos, extract subtitles, and generate static web pages with bilingual subtitles, timestamped key points, frame-matched knowledge cards, and quotes, enabling efficient learning from long-form content like Dan Koe's videos.
Qwenautomation pipelineknowledge management
0 likes · 9 min read

Tencent's Terminal-RSI framework uses recursive self-rewrite and policy refinement phases to train a Qwen3.8-27B terminal agent that achieves 86.52% on Terminal-Bench 2.1, surpassing GPT-6 Sol and Kimi K3, while demonstrating improved long-horizon task completion through credit assignment and reward design innovations.
Credit AssignmentLong-Horizon TasksQwen
0 likes · 13 min read

Renmin University introduces AEWM, an editing world model that classifies and revises noisy agent actions in real time, improving long-horizon task performance across search, terminal, and software engineering benchmarks without fabricating tool feedback.
AEWMAgent CorrectionEditAct
0 likes · 8 min read


java1234Oct 8, 2026 · Artificial Intelligence
This article details the architecture and implementation of an AI-powered OA system built with Spring AI 2.0, Spring Boot 4, and Vue 3, featuring intelligent leave form filling, approval progress querying, and RAG-based policy Q&A using Alibaba's Qwen and embedding models, with a four-role permission model, two-level approval engine, and Redis vector storage.
Alibaba BailianJWT AuthenticationMyBatis Plus
0 likes · 38 min read

Qwen researchers propose ID Balancing, a PID-based load control method for extremely sparse MoE models, achieving stable expert load with 768 experts and Top-3 routing, reducing MaxVio by over 50% and MinVio by 12% while maintaining competitive language modeling and downstream task performance.
DeepSeekMixture of ExpertsMoE
0 likes · 11 min read

JEMM-27B, a Qwen-based open-weight multimodal decision model, achieves 93.3% tool-calling accuracy and 54.6% screenshot web-task accuracy while cutting latency 31% versus Jev, enabling fast agent decisions via choice, judgment, and scoring APIs.
Agent ArchitectureApache 2.0JEMM-27B
0 likes · 9 min read

Within 48 hours of Jev's release, two open-source decision-model alternatives appear: Bespoke Nimble (9B, contrastive data curation, 90.12% accuracy) and Kev (0.5B LoRA on Qwen2.5, trains on a MacBook in 1h45m), both fully open and TypeSafe-compatible, demonstrating rapidly lowering barriers for System One-style routing and scoring models.
JevLoRA fine-tuningModel Evaluation
0 likes · 9 min read

Researchers from Nanjing University and University of Wollongong propose Harness Continual Learning (HCL), a framework where AI agents continuously adapt by evolving their system-level harness—prompts, memory, skills, tools, and routing—while keeping the base model frozen, demonstrating improved performance and controlled forgetting across diverse benchmarks.
ALFWorldAgent Continual LearningFrozen Model Adaptation
0 likes · 16 min read

Researchers discover S-Space, a stable low-dimensional spatial workspace in multimodal models that encodes object positions, but models confuse coordinate systems and fail to reliably rotate spatial representations, revealing a gap between having spatial representations and using them for reasoning.
Coordinate SystemsGPT-6 AstraInterpretability
0 likes · 15 min read

Long‑term LLM agents suffer from ever‑slowing execution and context poisoning because they continuously append every observation, action, and reasoning step to the prompt, but the SKILL.state approach replaces this growing history with a compact mutable state, dramatically cutting token usage while boosting accuracy and robustness across diverse benchmarks.
GeminiGemmaLLM Agents
0 likes · 11 min read

The newly open‑sourced Qwen 3.8‑27B model, built on the Qwen 3.5 architecture, expands its 27‑billion‑parameter foundation with stronger inference, programming, and autonomous‑agent abilities, achieving competitive benchmark scores and offering a step‑by‑step HyperAI tutorial for rapid deployment.
AgentLLMQwen
0 likes · 4 min read

The community quickly tested three newly released LLMs—DeepSeek V4 Pro, Qwen 3.8 Max, and Grok 4.6—across 3D scene generation, Flappy game creation, and airplane‑animation tasks, comparing quality, speed, and cost to reveal each model’s strengths and trade‑offs.
AIDeepSeekGrok
0 likes · 5 min read

The article analyzes Apple’s move to integrate Alibaba’s Qwen large‑model into Siri for the Chinese market, explaining the regulatory, ecosystem and strategic reasons behind the partnership, and discussing how this signals a shift from benchmark hype to system‑level AI orchestration.
AI ecosystemAppleQwen
0 likes · 4 min read

The article dissects recent Agentic RL breakthroughs—including GLM‑5.2’s shift from GRPO to critic‑based PPO, Qwen’s multi‑dimensional verification system, the generative‑critic GenAC, and the on‑policy skill‑distillation method OPID—showing how each tackles long‑trajectory credit assignment, reward hacking, and scalable evaluation across software‑engineering, front‑end, and real‑world tasks.
GLM-5.2Generative CriticLLM Evaluation
0 likes · 36 min read

Claude Code only understands Anthropic‑format endpoints, but by adjusting three environment variables you can route it to domestic models such as Kimi K3 or Alibaba Qwen, with detailed step‑by‑step instructions for direct env‑var changes, settings‑file configuration, and GUI or CLI tools, plus known limitations.
Anthropic APIClaude CodeKimi
0 likes · 12 min read

Alibaba Mama AI Infra and AMD collaborated to integrate SGLang‑ATOM plugin mode for Qwen 3.5/3.6 models on MI308X GPUs, detailing multimodal, GDN, KV‑cache, and parallel‑strategy optimizations that achieve low‑latency, high‑throughput inference suitable for complex advertising workloads.
AMD ATOMGPU optimizationLLM inference
0 likes · 18 min read

DeepSeek’s open‑source DSpark applies speculative decoding to its V4 Flash and Pro models, delivering 51%‑400% inference throughput gains that vary by task, while also supporting other models such as Gemma and Qwen, positioning it as a versatile, cross‑model acceleration solution.
AI Inference AccelerationDeepSeekGemma
0 likes · 6 min read

The author built a zero‑cost internal coding assistant using iFlytek's free Qwen3.6‑35B‑A3B and Qwen3.5‑35B‑A3B models, explains why these models were chosen over alternatives, provides a nine‑step guide to claim the free MaaS token quota, shares ready‑to‑run Python code, and reports real‑world performance across code generation, long‑document parsing, and multi‑turn conversations, while also outlining suitable user groups and an optional enterprise Token Plan.
APICode AssistantMaaS
0 likes · 12 min read

The paper introduces NEO‑ov, a native vision‑language model that discards external visual encoders, feeding raw pixels directly into a unified transformer, and demonstrates competitive performance on image, multi‑image, and video tasks—including fine‑grained perception and spatial reasoning—while outlining its three‑stage training pipeline and current limitations.
Qwenbenchmarkmultimodal
0 likes · 13 min read

This guide explains how to configure Codex’s App, CLI, and SDK to use third‑party large language models such as Qwen, MiniMax, DeepSeek, and local Ollama/LM‑Studio models by adding provider entries to the ~/.codex/config.toml file and selecting the appropriate wire_api (chat or responses).
CodexLM StudioMiniMax
0 likes · 8 min read

Tencent has contributed $20 million to Lin Junyang’s newly founded AI Lab, bringing the post‑money valuation to roughly $2 billion, while the founder—formerly Alibaba’s youngest P10 technical executive and key figure behind the Qwen model series—already begins looking for a follow‑up funding round.
AI fundingAlibabaLin Junyang
0 likes · 4 min read

The Qwopus3.5‑9B‑Coder model, fine‑tuned for agentic coding, tool calling and logical reasoning, offers three formats (Safetensors, GGUF, GGUF+MTP), runs on a 16 GB Mac mini via LM‑Studio, achieves up to 35% throughput gain with MTP, scores 85 on HermesAgent‑20, 100 on ToolCall‑15, and 53.89% on SWE‑bench, matching Claude Opus 4.6 in a 31‑tool adversarial test while highlighting its training tricks and current limitations.
Agentic CodingLLM benchmarkQwen
0 likes · 11 min read

This article explains how developers can bypass token‑cost barriers by using iFlytek’s MaaS platform to obtain free, unlimited‑token access to the Qwen 3.6‑35B‑A3B model, details the model’s MoE architecture and benchmark performance, and provides a complete Java integration tutorial with code samples and practical use‑case suggestions.
AIAPIJava
0 likes · 16 min read

This step‑by‑step guide shows how to install CC Switch v3.16.0, add DeepSeek or Qwen as a provider, enable local routing, and switch Codex to these third‑party large language models, preserving the original Codex experience while reducing API costs.
AI modelsCodexDeepSeek
0 likes · 6 min read

This article analyzes the hidden pitfalls of Qwen's official Jinja chat template, explains how the community‑maintained Qwen‑Fixed‑Chat‑Templates v19 fixes rendering errors, KV‑Cache loss, token waste and agent dead‑locks, and provides step‑by‑step installation instructions for LM Studio, llama.cpp, vLLM and MLX.
KV CacheLLMQwen
0 likes · 10 min read

The Qwopus 3.6‑27B‑v2 model reconstructs full step‑by‑step reasoning from compressed Claude outputs using a Trace‑Inverter, creates two high‑quality SFT datasets, and achieves 35.9% token savings, a 2.57‑point accuracy gain on MMLU‑Pro, 75.25% success on SWE‑bench, while running on a single consumer‑grade RTX 5090.
GGUFMMLUQwen
0 likes · 11 min read


SuanNiMay 19, 2026 · Artificial Intelligence
Qwen 3.7‑Max‑Preview secures the 13th spot worldwide and the top position among Chinese models, while Qwen 3.7‑Plus‑Preview ranks 16th in vision, highlighting an accelerated release cadence, deeper technical depth across sub‑tasks, and a shift in China’s large‑model competition toward ecosystem control.
AI competitionChina AIQwen
0 likes · 9 min read

The Qwen 3.7‑Max and Qwen 3.7‑Plus preview models debut with top‑15 global rankings in Arena, the only Chinese models in text and vision leaderboards, while a timeline analysis shows the Qwen series accelerating from 4‑6‑month releases to a 2‑3‑month cadence and introducing dense and MoE variants up to 235 B parameters.
AI benchmarkChinese AIModel Iteration
0 likes · 6 min read

vLLM tops the Artificial Analysis ranking by delivering the highest throughput for DeepSeek V3.2, Qwen 3.5 397B, and MiniMax‑M2.5 on identical NVIDIA Blackwell Ultra hardware, thanks to extensive kernel‑fusion optimizations that remain in the main branch.
DeepSeekLLM inferenceQwen
0 likes · 7 min read

Researchers at Claude applied the MSM (mid‑training) approach to Qwen models, inserting a value‑document pre‑training phase before alignment fine‑tuning, which reduced misalignment rates from 68%/54% to 5%/7% and cut required fine‑tuning data by 40‑60×, demonstrating superior generalization when combined with standard alignment.
AI AlignmentMSMQwen
0 likes · 6 min read

Qwen‑Scope adds a sparse autoencoder (SAE) to the Qwen‑3.5‑27B model, exposing a top‑K 50‑feature, residual‑stream hook across all 64 layers for interpretability, controllable generation, data analysis, and training diagnostics, while detailing installation, usage, and practical trade‑offs.
InterpretabilityQwenSAE
0 likes · 11 min read


SuanNiApr 22, 2026 · Artificial Intelligence
Alibaba’s newly released open‑source Qwen 3.6‑27B dense model, with 27 billion parameters, beats its 397 billion‑parameter predecessor across a suite of code‑generation and multimodal benchmarks, while offering easier deployment thanks to its pure‑dense architecture and native image‑video‑text capabilities.
Dense ArchitectureQwenbenchmark
0 likes · 5 min read

The newly released Qwen3.6-27B dense multimodal model, at just 27 B parameters, surpasses the 397 B flagship on most encoding benchmarks, offers up to 1 M token context, supports FP8 quantization, and can be deployed locally via vLLM, SGLang or Transformers with modest hardware.
27BDense ModelFP8
0 likes · 12 min read


SuanNiApr 21, 2026 · Artificial Intelligence
The article analyzes Qwen3.6‑35B‑A3B’s MoE architecture, showing how its 30 B active parameters outperform larger dense models across programming, agent, and multimodal benchmarks, and examines the flagship Qwen3.6‑Max‑Preview’s substantial gains in world knowledge, instruction following, and third‑party rankings.
AI-evaluationMixture of ExpertsQwen
0 likes · 5 min read

In the last 24 hours the AI landscape shifted not by a handful of new model releases but by two converging fronts—model‑level advances in agentic coding and product‑level moves that turn models into usable work systems—signaling deeper changes in competition and industry impact.
AI modelsAgentic CodingClaude
0 likes · 14 min read

The article analyzes how reward‑shaping techniques can shorten the chain‑of‑thought outputs of Qwen 30‑parameter series models by 20‑40% while preserving or slightly improving performance on AIME‑25 and out‑of‑distribution benchmarks, and it details the experimental design, strategic considerations, and practical insights behind this efficient reasoning approach.
QwenReinforcement LearningReward Shaping
0 likes · 16 min read

Researchers introduce Latent Reasoning Tuning (LRT), a lightweight inference network that encodes explicit reasoning chains into fixed‑length latent vectors, eliminating thousands of decoding steps; experiments reveal substantial redundancy in traditional chains and demonstrate that LRT achieves faster, more accurate inference and outperforms existing efficient reasoning methods.
DeepSeekHybrid ReasoningQwen
0 likes · 10 min read

An in‑depth 2026 comparative review of Alibaba’s Qwen series (Qwen2.5, Qwen3, Qwen3.5) evaluates architecture, performance, speed and VRAM usage on Mac Studio, ranks each variant, and provides concrete model‑selection guidance for different memory configurations, highlighting the MoE‑based Qwen3.5 as the optimal choice.
AI performanceMac StudioMoE
0 likes · 9 min read

This guide presents an end‑to‑end solution for creating a test‑focused AI agent that indexes project code and defect data, integrates a large language model via LangChain, generates compliant Pytest cases, parses Allure reports, and offers deployment tips for seamless PyCharm integration.
AI AgentAllureLangChain
0 likes · 13 min read

In March 2026, OpenRouter recorded 9.55 trillion tokens consumed weekly, with Chinese models occupying six of the top‑10 slots, Qwen surpassing 1 billion downloads, and cost advantages that let domestic LLMs outpace U.S. counterparts in both performance and price.
AI CostChinese LLMsMiniMax
0 likes · 9 min read

This guide explains why selecting the correct AI model is crucial for OpenClaw, outlines the problems caused by a mismatched model, and provides step‑by‑step commands to list providers, authenticate, set defaults, and customize routing for models such as Qwen, OpenAI, Anthropic, and Gemini.
AI agentsAuthenticationOpenClaw
0 likes · 5 min read

On March 4, 2026, Alibaba’s youngest P10 technical leader Lin Junyang announced his resignation with a nine‑word tweet, just hours after releasing four Qwen 3.5 models that earned Elon Musk’s praise, while two other core researchers also left, leaving the future of China’s leading open‑source AI flagship uncertain.
AI talent turnoverAlibabaChina AI
0 likes · 9 min read

This article walks through building a Python pipeline that uses DeepSeek and Alibaba Qwen to generate AI‑written articles, creates cover images, and automatically saves them as drafts in a WeChat public account, with detailed environment setup, client implementations, fallback strategies, and deployment tips.
AIDeepSeekPython
0 likes · 26 min read

This article reviews the three open‑source Qwen 3.5 models—including a 35B MoE, a 122B MoE, and a 27B dense version—detailing their parameter layouts, core attention designs, context length, inference performance, hardware requirements, and provides step‑by‑step code examples for loading them with Hugging Face Transformers and vLLM.
AIMoEQwen
0 likes · 10 min read

This article explains how Chaterm's Agent Skills, powered by the Qwen large model, let you package operational expertise into reusable, executable skills that automatically diagnose and fix issues, turning a manual 20‑minute troubleshooting process into a three‑minute AI‑driven workflow.
AI OpsChatermDevOps
0 likes · 14 min read

The open‑source Qwen‑Image‑Edit‑2511‑Multiple‑Angles‑LoRA model can reconstruct images from 96 preset camera poses, letting users adjust distance, pitch and yaw to generate realistic multi‑angle views, with step‑by‑step usage instructions, example results, practical applications, and noted limitations.
AIQwenimage editing
0 likes · 6 min read

This article proposes an ASR‑driven pipeline that creates high‑quality AI‑reply vs. human‑like reply pairs, trains a rewrite model and a reward model, and uses GRPO reinforcement learning to generate natural, helpful, and less AI‑sounding responses in digital‑human live streaming, achieving 92% accuracy and 97% helpfulness while improving user experience.
ASR dataLLMQwen
0 likes · 20 min read

Ollama v0.14.0 now supports the Anthropic API, letting you run Claude Code locally with open‑source models like Qwen or Llama without an API key, network, or cost, and the article provides a step‑by‑step setup, SDK examples, and an objective assessment of the approach.
Anthropic APIClaude CodeOllama
0 likes · 7 min read

In 2025 China's AI large‑model arena is sharply fragmenting, with ByteDance's Doubao leading user activity, DeepSeek dominating technical and international influence, and Alibaba's Qwen carving a unique full‑stack strategic edge, each pursuing distinct paths in technology, product and ecosystem competition.
AIChinaDeepSeek
0 likes · 11 min read

This guide shows how to use the Qwen large language model to automatically generate pytest‑style API test scripts from an OpenAPI specification, covering setup, prompt design, model invocation, code extraction, and execution steps.
AI code generationAPI TestingOpenAPI
0 likes · 9 min read

NeuTTS‑Air, an open‑source on‑device text‑to‑speech model built on a 0.5B Qwen LLM and NeuCodec, reaches SOTA among open models, runs entirely on CPU, supports 3‑second voice cloning, and comes with a step‑by‑step tutorial for deployment on edge devices.
NeuCodecNeuTTS-AirOn-device Inference
0 likes · 5 min read

Learn how to rapidly create a full-featured AI-driven Flutter mobile application that generates and edits 3D figurine images using Qoder's code generation, Alibaba Cloud ADB Supabase for backend-as-a-service, and the Qwen Image Edit model, all without building a traditional backend.
AIEdge FunctionFlutter
0 likes · 10 min read

Learn how to leverage Spring AI's prompt template feature to create flexible, variable-driven queries, and implement backend services using DeepSeek and QWEN models for dynamic travel recommendations, complete with code examples for interfaces, service implementations, and controller routing.
DeepSeekJavaPrompt templates
0 likes · 7 min read

The article provides a comprehensive technical walkthrough of Qwen’s large‑model series, covering data preparation, tokenization, model tweaks, training settings, RLHF pipeline, Code‑Qwen specifics, Qwen2 and Qwen3 architectural changes, scaling‑law experiments, and detailed source‑code analysis with illustrative diagrams.
MoEQwenRLHF
0 likes · 7 min read

This article reviews the latest updates from major LLM providers—DeepSeek V3’s parameter boost and longer context, Qwen2.5‑Omni’s open‑source multimodal 7B model, Google Gemini 2.5 Pro’s 1 M‑token window and multimodal prowess, and OpenAI GPT‑4o’s image generation and reduced pricing—highlighting technical specs, capabilities, and availability.
DeepSeekGPT-4oGemini
0 likes · 9 min read

This tutorial explains how large language models can use function calling to access real‑time data, walks through setting up a Flask endpoint, demonstrates integration with Mistral Small, Llama 3.2‑1B, and Qwen models, and provides complete Python code examples for end‑to‑end execution.
APIFunction CallingLLM
0 likes · 10 min read

This article recounts a three‑stage interview for a multimodal large‑model position at Taobao's Search AI division, detailing typical questions on CLIP, LoRA, BLIP, Qwen‑VL, Transformer fundamentals, RLHF, and coding challenges, and offers insights on what interviewers focus on.
AICLIPInterview
0 likes · 5 min read

This article details how a 7B parameter language model, fine‑tuned with DeepSeek's GRPO reinforcement‑learning algorithm and a carefully crafted multi‑component reward system, learned to solve Sudoku puzzles without any cold‑start data, outperforming a comparable 3B model and revealing key insights for structured reasoning tasks.
AI TrainingGRPOQwen
0 likes · 15 min read

This guide walks you through fine‑tuning a large language model without any coding by using LLaMA‑Factory, covering LoRA fundamentals, environment setup, dataset creation, parameter configuration, training, loss monitoring, model export, and a quick evaluation on the Qwen2.5‑0.5B model.
AnacondaLLaMA-FactoryLoRA
0 likes · 15 min read

Alibaba has released the open‑source Qwen QwQ‑32B model, a 32‑billion‑parameter LLM that matches DeepSeek‑R1's performance while being deployable on consumer‑grade GPUs, and the announcement is accompanied by extensive promotional offers for AI‑related products and services.
AI benchmarkAlibabaQwen
0 likes · 7 min read

Alibaba’s newly released 32‑billion‑parameter QwQ‑32B model matches the performance of 670‑billion‑parameter rivals like DeepSeek‑R1, integrates agent‑based reasoning, runs on consumer hardware, and has sparked strong open‑source community adoption, as shown by benchmark results and download statistics.
AgentAlibabaQwen
0 likes · 6 min read

This tutorial explains how to build a distributed reinforcement‑learning pipeline using the GRPO algorithm, covering data preparation, evaluation and reward functions, multi‑GPU DataParallel implementation, and full fine‑tuning of the Qwen2.5‑1.5B‑Instruct model with PyTorch, FlashAttention2 and Weights & Biases.
AIGRPOPyTorch
0 likes · 10 min read

This article explains the FlashAttention algorithm, its memory‑efficient tiling and recomputation techniques, and how enabling the flash_attn flag dramatically speeds up Qwen‑series large models while outlining hardware, software requirements and potential trade‑offs.
AttentionFlashAttentionGPU optimization
0 likes · 8 min read

Apple’s new Apple Intelligence platform, unveiled at WWDC24, will incorporate Alibaba’s Qwen 2.5 Max model to create China‑specific AI features for iPhone users, with a custom dataset and regulatory submission, marking a shift from overseas ChatGPT reliance to a domestic partnership.
AIAlibabaApple
0 likes · 3 min read

This guide walks through the recent DeepSeek R1 and Qwen2.5 releases, explains why running LLMs on a PC or phone saves compute and token costs, shows step‑by‑step installation of Ollama, building a Spring AI Alibaba chat app, and scaling the solution with the Higress cloud‑native API gateway for production use.
DeepSeekHigressLLM
0 likes · 11 min read


JavaEdgeDec 1, 2024 · Artificial Intelligence
QwQ‑32B‑Preview, an experimental AI model from the Qwen team, showcases strong reasoning in math and programming while facing challenges like language switching, inference loops, safety concerns, and variable capabilities across domains, with benchmark scores ranging from 50% to over 90% on tests such as GPQA, AIME, MATH‑500, and LiveCodeBench.
AI benchmarkLLMModel Evaluation
0 likes · 7 min read

The author chronicles the challenges and solutions of running large language models locally using Ollama, experimenting with cloud GPUs on Google Colab, managing Python dependencies through Docker, and ultimately fine‑tuning a small Qwen model, providing a practical guide for AI enthusiasts.
DockerGoogle ColabLLM
0 likes · 6 min read

This tutorial walks through setting up a local k3d cluster, installing Higress, and using its AI plugins—including AI Proxy, AI JSON formatter, AI Agent, and AI Statistics—to integrate and observe Alibaba Cloud's Qwen large language models across various use cases such as weather and flight queries.
AI gatewayAI pluginsHigress
0 likes · 30 min read

This guide shows how to use free AI translation tools—Immersive Translate and OpenAI Translator—together with the Higress cloud‑native AI‑proxy plugin, configuring Docker, model mappings, and custom dictionaries to efficiently translate computer‑science textbooks like Rust and Crafting Interpreters, while comparing machine and human translations.
AI translationDockerHigress
0 likes · 11 min read

This article walks through setting up a cloud‑native test environment using the open‑source NextChat UI and Higress API gateway to let Qwen‑2.5 masquerade as GPT‑4o, enabling a side‑by‑side comparison of their responses while showcasing Higress’s streaming, hot‑update, and security features for AI workloads.
AI gatewayDockerGPT-4o
0 likes · 8 min read

Qwen1.5‑MoE‑A2.7B is a 2.7 billion‑parameter Mixture‑of‑Experts model that delivers performance comparable to leading 7 billion‑parameter LLMs while cutting training cost by 75% and boosting inference speed by 1.74×, and the article details its architecture, benchmarks, efficiency analysis, and deployment steps.
MoEQwenTraining Efficiency
0 likes · 13 min read
