#active parameters
1198 related articles
·2 min
MiniMax M3 Local Deployment: 23B Active Parameters ≠ 23GB VRAM — A Complete Guide to MoE Architecture
MiniMax M3 is a 428B MoE model. Its 23B active parameters reflect per-token compute, not VRAM needs. Learn the MoE trade-offs, quantization options, and deployment paths to avoid the most common misconception.
Read more →
Tech Frontiers·2 min
LFM2.5-8B-A1B: A MoE Model with 1.5B Active Parameters Delivering 4x Its Weight Class Performance
Liquid AI releases LFM2.5-8B-A1B, a MoE model with 8B total params but only 1.5B active, matching 6B-class models in tool calling. Supports 128K context, local deployment, multilingual, with SGLang Day-0 support.
Read more →
·2 min
Why Top AI Startups Are No Longer Publishing Research: The Deep Reasons Behind the Shift from Openness to Secrecy
Exploring why top AI startups shifted from open research to secrecy, analyzing how commercial competition and talent pressure drive this change, and its impact on academia, innovation, and open source.
Read more →
·4 min
Infrastructure Architecture for Agent Applications: Four Core Patterns Explained
A deep dive into infrastructure architecture patterns for production-grade Agent applications, covering state persistence, sandbox isolation, LLM observability, and cost control.
Read more →
·4 min
Infrastructure Architecture for Agent Applications: Four Core Patterns Explained
Deep dive into infrastructure architecture patterns for production-grade Agent applications, covering state persistence, sandbox isolation, LLM observability, and cost control.
Read more →
·3 min
Ollama Template Silently Overridden After GGUF Import: A Troubleshooting Guide for Custom TEMPLATE Being Ignored
Analysis of Ollama silently overriding custom TEMPLATE with GGUF-embedded templates after import. Learn to verify active templates via /api/show to prevent evaluation failures.
Read more →
·3 min
TokenTown: A Visual Approach to Understanding How LLMs Predict the Next Token
TokenTown is an open-source visualization project that intuitively presents the internal token prediction process of LLMs using a town metaphor. Learn its design philosophy and educational value.
Read more →
·3 min
What Are AI Model Weights, Really? The Deep Learning Truth Revealed by a Meme
Starting from a viral Reddit meme, we dive deep into AI neural network weights — what they are, why they can't be read visually, and how open weights drive technological democratization.
Read more →
·3 min
Production-Grade LLM Inference Deployment: A Practical K8s Handbook
Complete guide to deploying production-grade LLM inference on Kubernetes, covering GPU scheduling, vLLM engine selection, autoscaling, observability, and cost optimization.
Read more →
·3 min
Two Highly Effective AI Debugging Prompts: Cracking YOLOv8 and OpenCV Challenges
In-depth analysis of two battle-tested AI debugging prompts for diagnosing YOLOv8 training mAP collapse and OpenCV RTSP stream corruption, revealing structured debugging prompt design patterns.
Read more →
·3 min
Autonomous Combat Tanks: Technical Implementation and Ethical Boundaries of AI Militarization
In-depth analysis of AI autonomous combat tanks: reinforcement learning training, environmental perception, decision engines, global military AI competition, and the ethical dilemmas of lethal autonomous weapons systems.
Read more →
·3 min
Is LLM Inference Actually Profitable? Breaking Down the Cost Structure and Business Logic of Large Model Inference
A deep dive into LLM inference cost structure and profitability models—from GPU throughput, MoE architecture, and KV Cache to scale effects—revealing the business logic behind API price wars.
Read more →
·4 min
9 Hours, 4 Million Tokens: AI Creates a Stunning WebGPU Snow Rendering Masterpiece with Claude Code
Reddit user Noniv used Claude Code with the Opus model to build SNOWFLOW, a stunning WebGPU real-time snow renderer, in just 9 hours using 4 million tokens. A deep dive into prompt strategy, snow shaders, and performance optimization.
Read more →
·3 min
Altman Warns of AI Monopoly Risk: A Few Companies Controlling AI Would Be Extremely Dangerous
OpenAI CEO Sam Altman warns that AI controlled by a few companies would be very dangerous. We analyze the real threats, his complex motivations, and paths to breaking AI monopoly.
Read more →
·3 min
Altman Warns of AI Monopoly Risk: A Few Companies Controlling AI Would Be Extremely Dangerous
OpenAI CEO Sam Altman warns that AI controlled by a few companies would be extremely dangerous. This article analyzes the real threats of AI monopoly, Altman's complex motivations, and paths to breaking concentration through open source, compute democratization, and regulation.
Read more →
·3 min
The Era of AI Capability Overhang: Why You Need to Reset Your Ambition Every 3 Months
Understanding Capability Overhang in the AI era: when model capabilities far exceed application imagination, how teams should reset feasibility boundaries quarterly to avoid ceding advantages to competitors.
Read more →
·4 min
Project Rai-chan Technical Breakdown: A Complete Guide to Building a Fully Local AI Companion
Deep dive into Project Rai-chan's tech stack: Ollama+Gemma local LLM, Unity rendering, VOICEVOX speech synthesis, and more — exploring the technical path for local AI companions.
Read more →
·4 min
A Guide to Full-Stack Machine Learning Project Repository Structure: From Chaos to Convention
A detailed guide to organizing full-stack ML project repositories, covering directory structure design, data-code separation, and externalized configuration to help ML developers move from experimental code to production-grade engineering standards.
Read more →
·4 min
Full-Stack Machine Learning Project Repository Structure Guide: From Chaos to Convention
A detailed guide to organizing full-stack ML project repositories, covering directory structure design, data-code separation, and configuration externalization to help ML developers move from experimental code to production-grade engineering.
Read more →
·2 min
Running Manim in the Browser: Rebuilding 3Blue1Brown's Animation Engine with WebGPU
Explore how WebGPU brings 3Blue1Brown's Manim math animation engine to the browser, enabling zero-install, real-time mathematical visualization with interactive education potential.
Read more →