@akokoi1: As long as the Skills are well-written, they are not model-dependent. My edulab Skills (generating 3D geometry problem-solving demonstrations based on questions) successfully completed the task on Qwen3.8 27B UD-IQ4_XS. Computer setup: CPU: 12400F GPU: RTX3080 20G…
Summary
The user successfully used edulab Skills to generate 3D geometry problem-solving demonstrations on the Qwen3.8 27B model and discussed performance and optimization needs under specific hardware configurations.
View Cached Full Text
Cached at: 08/24/26, 01:45 AM
As long as your Skills are well-written, they don’t really depend on the model. The Skills I developed—edulab Skills (which generates 3D geometric problem-solving demonstrations based on questions)—successfully completed the task on Qwen3.8 27B UD-IQ4_XS.
Computer Configuration: CPU: 12400F GPU: RTX3080 20G RAM: DDR4 3200 32G
Agent Tool: WorkBuddy
Model context window: 128k. Deep thinking mode enabled. Speed: 35 tok/s. Task completion time: 16 minutes. Context usage: 83k.
This Skill still needs optimization—aiming to reduce generation time while maintaining quality.
WY (@akokoi1): Finally ordered a modified 20G 3080. The client has local data privacy requirements, so I’m starting with a basic GPU to test Qwen3.8 27B’s performance limits. This card should be sufficient for testing local data cleaning capabilities.
Similar Articles
@zhixianio: After receiving the new machine, I began an 'ascetic' practice of forcing myself to use local models for common tasks. I thought it would be painful, but both speed and quality greatly exceeded my expectations: Model: Qwen3.6-35B-A3B-oQ6-fp16-mtp, Running: oMLX, with N…
The author uses the Qwen3.6-35B-A3B model and oMLX tool on the new local machine for daily tasks, finding that both speed and quality far exceed expectations, even outperforming remote LLMs in PA and coding scenarios, demonstrating a significant improvement in on-device AI capabilities.
@servasyy_ai: https://x.com/servasyy_ai/status/2091416214283379123
This article details the local deployment guide for the Qwen3.8 27B model, covering two routes for Mac and Nvidia graphics cards, and provides real-world performance data to help users run this model on consumer-grade hardware.
@Xudong07452910: Let the model set its own problems, solve its own problems, and train itself — the biggest fear is learning incorrect problems along with the correct ones. This paper by the Qwen team proposes Skill Self-Play, adding a continuously updated skill library to the model's self-evolution. There are three roles in training: The Proposer generates tasks that are just challenging enough based on the skills...
The Qwen team proposes the Skill Self-Play framework, which significantly improves model capabilities on tool-calling and reasoning tasks through the collaboration of Proposer, Solver, and a dynamic skill controller in self-play.
@akokoi1: The first skill of this AI teaching tool, "3D Geometry Problem Solving," has been released, supporting text-based problem solving, image-based problem solving, and random question generation. After testing mainstream models, deepseek-v4-pro offers the best cost-performance ratio at 0.2-0.3 yuan per question, with accuracy on some questions even surpassing GPT5.5.
The first skill of the AI teaching tool, "3D Geometry Problem Solving," has been released, supporting text-based problem solving, image-based problem solving, and random question generation. Tests show that deepseek-v4-pro offers high cost-performance, with accuracy on some questions even surpassing GPT5.5.
@MinLiBuilds: The GGUF version of Qwen3.8-Flash-Next, an 111GB model, surprisingly runs on dual RTX 4090 with 96GB VRAM. This model is truly bizarre. My setup: 2× RTX 4090 48G = 96GB VRAM, 62GB RAM, using Unsloth's …
Successfully ran the 111GB Qwen3.8-Flash-Next GGUF model on dual RTX 4090 with 96GB VRAM, achieving a generation speed of 66 tokens/s through PLE and sparse attention optimizations, with detailed explanations of quantization and runtime configuration.