#Qwen

共 31 篇带此标签

31 articles Newest 2026-11-17
Archive · #Qwen 11 articles
LLMs

Qwen4-Max: Agent Selection Goes Beyond Benchmarks

Qwen4-Max reached the Hacker News front page with 483 points and 197 comments, but Agent teams should evaluate more than model rankings. A reproducible baseline of 30 tasks, each run five times, reveals the real cost, latency, and recoverability of each option.

2026-11-16 Source 2026-11-16 8 min 115 views
LLMs

Choosing Qwen3.8-27B Quantization by VRAM

A practical framework for comparing Qwen3.8-27B quantizations across VRAM limits, using Unsloth's Dynamic v3.1 guidance and reproducible benchmarks for speed, memory use, output quality, and framework compatibility.

2026-11-15 Source 2026-11-14 8 min 115 views
Case Studies

Qwen 3.8 27B Inspects Model Weights Locally

Simon Willison ran Qwen 3.8 27B on his laptop to inspect another model’s weights. The experiment offers a practical workflow for connecting a local LLM to reproducible statistics, hypothesis generation, and verification without treating generated explanations as causal proof.

2026-11-04 Source 2026-11-03 8 min 115 views
Products

Astra Weight Editing: Agents That Modify Models

Astra turns Qwen model weights into a programmable interface that Agents can read and modify. This article examines the implementation, reproducibility barriers, and how weight editing differs from prompting, fine-tuning, and tool use.

2026-11-03 Source 2026-11-02 9 min 115 views
Frameworks

Five Qwen Agents Beat One Stronger Model

Memdoor used five Qwen3.8-27B instances to outperform Fable 5 on LiveCodeBench Hard. Its team-based workflow offers a practical blueprint for multi-Agent orchestration, shared memory, and result revision.

2026-11-01 Source 2026-10-31 6 min 115 views
Case Studies

Can Qwen3.5 Really Design a RISC-V CPU?

Qwenputer uses Qwen3.5 to build an executable, deterministic RISC-V CPU. The project offers a useful test of AI-assisted hardware engineering, while also revealing the limits of its current verification.

2026-10-29 Source 2026-10-29 5 min 115 views
Tools

GVS5H Review: Five Qwen3.8-27B Models Match Fable 5

A Hacker News discussion reports that five Qwen3.8-27B models matched Fable 5 through the GVS5H harness. A rigorous comparison requires auditing the benchmark protocol and controlling for prompts, tools, retries, and scoring.

2026-10-25 Source 2026-10-25 7 min 115 views
Opinion

DeepSeek 4.1 Flash Lands Quietly. The Open-Source Hype Is Exhausted.

DeepSeek's latest release drew 212 points and 181 comments on Hacker News — respectable for most AI labs, underwhelming for the company that once moved Nvidia's stock. The numbers tell the story: open-source model releases have stopped triggering panic, and the disruption narrative may have run its course.

2026-10-09 Source 2026-10-08 4 min 115 views