279 related articles

Running Kimi K3 with 29GB RAM at just 0.5 tok/s. An in-depth analysis of extreme quantization techniques, performance trade-offs, and the impossible triangle of local LLM deployment.

Running Kimi K3 with 29GB RAM at just 0.5 tok/s. A deep analysis of extreme quantization techniques, performance trade-offs, and the impossible triangle of local LLM deployment.

Deep analysis of how cross-cloud GPU preemption migration technology helps MLOps teams cut 40% of compute costs through predictive telemetry, cross-cloud state migration, and compute arbitrage.

Deep analysis of the SDL_GPU single-header 2D graphics library covering its design philosophy, GPU acceleration principles, use cases, and technical trade-offs.

Speech To Markdown is a free macOS/iOS app that converts voice to structured Markdown notes using local LLMs. Fully offline, no API keys needed, with global hotkey dictation.

Analyzing real LLM inference costs: from B200 GPU compute gains, vLLM framework optimization to MTP multi-token prediction, explaining why serving costs are widely overestimated.

OpenAI's internal model GPT-5.6 reportedly autonomously rewrites production kernels, achieving ~20% service cost reduction. Deep analysis of this AI recursive self-optimization event's technical plausibility and industry impact.

OpenAI's internal model GPT-5.6 reportedly autonomously rewrote production compute kernels, achieving ~20% cost reduction. Deep analysis of this AI recursive self-optimization event's technical plausibility, industry impact, and key questions.

Local LLM crashing in Agent frameworks? The issue may be num_gpu set too high. Learn what num_gpu really controls (GPU layer offloading, not GPU count) and how to tune it for stable Agent performance.

Complete guide to deploying production-grade LLM inference on Kubernetes, covering GPU scheduling, vLLM engine selection, autoscaling, observability, and cost optimization.

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.

Moonshot AI open-sources FlashKDA, providing high-performance CUDA kernels for Kimi Delta Attention. Learn about its technical principles, performance gains, and value for long-context training and inference acceleration.

Moonshot AI open-sources FlashKDA, providing high-performance CUDA kernels for Kimi Delta Attention. Explore its technical principles, performance gains, and value for long-context training and inference.

Deep dive into the maderix/ANE GitHub project that reverse engineers Apple's private APIs to enable neural network training on the Apple Neural Engine, exploring its technical approach, efficiency gains, compliance risks, and implications for on-device AI.

Deep dive into Kimi Delta Attention (KDA): from standard Softmax attention's quadratic bottleneck through linear attention, Delta Rule, and gated decay mechanisms — the complete evolution explained.

A deep dive into Kimi Delta Attention (KDA): tracing the evolution from quadratic Softmax attention through linear attention, Delta rules, and gated decay mechanisms, with insights on associative memory and hardware optimization.

Facing GPU fragmentation on edge devices, the PostSlate team used ncnn's Vulkan backend for cross-platform ML inference, achieving 10× speedup on RTX 4070 with half the model size and zero runtime installation.

Moonshot AI launches Kimi K3 with 2.8 trillion parameters and 1M token context. Google delays Gemini 3.5 Pro, AI coding tools upgrade collectively as competition shifts to coding and Agent capabilities.

July 24 AI news: Black Forest Labs launches Flux 3 multimodal model, Kimi K3 lags in US-UK gov tests, Alibaba Qwen tops TTS rankings, Etched raises $300M, AMD unveils MI430X.

Moonshot AI releases Kimi K3, a 2.8 trillion parameter open-source model using MoE architecture that tops the global frontend coding arena at under $1 per task, beating GPT and Claude.