171 related articles

The MELTing Point paper is the first to evaluate mobile LLM performance in real user scenarios, covering iPhone, Samsung, Pixel and more, testing TinyLlama, Mistral-7B and others—revealing GPU inference gains, 47°C heat warnings, and prefill-decode disaggregation.

Deep analysis of Google's AI full-stack strategy: from custom TPU chips and system software frameworks to Gemini models and applications, examining how vertical integration delivers performance, cost, and autonomy advantages.

The Miles team and AMD announce the full port of DeepSeek-V4 Flash RL training to AMD Instinct MI355X GPUs on ROCm, boosting AIME pass@1 from 0.39 to 0.49—a milestone for compute ecosystem diversity.

In-depth Grok 4.5 hands-on review: priced at a fraction of Opus 4.8, twice the token efficiency of peers, and coding ability in the top tier. A real-project breakdown of its strengths, highlights, and shortcomings.

OpenAI's GPT Live full-duplex voice model, Grok 4.5 coding model with Cursor, and ByteDance's Seedream 5.0 Pro image generation launched together. A deep dive into three AI releases.

Meta launched an enterprise AI agent, with Zuckerberg claiming it can "run your entire business." This article explores the commercial value of AI agents, the hidden risks of data ownership, and how different businesses can balance efficiency with data sovereignty.

A hands-on comparison of 6 open-source LLMs (DeepSeek, Qwen3, Zhipu GLM, Kimi K2, MiniMax M3, Tencent Hunyuan 3) for on-premise deployment—covering hardware cost, inference efficiency, and deployment difficulty.

Deep dive into DeepSeek-V4: 1.6T-parameter MoE, CSA+HCA hybrid attention, MHC & MUON optimizer. Inference FLOPs drop to 27% of V3.2, redefining open-source LLM SOTA.

NVIDIA TensorRT now supports multi-device inference via pipeline and tensor parallelism, distributing large models across multiple GPUs to break through single-card memory limits.

Databricks tech lead Sandy shares a five-pillar framework for production-grade AI Agents—evaluation, observability, data foundation, orchestration, and governance—with a £85K retail banking failure case to bridge the demo-to-production gap.

A Databricks expert breaks down the complete methodology for taking AI Agents from demo to production, covering the five pillars of evaluation, observability, data foundation, multi-Agent orchestration, and AI governance, with a real eight-week banking chatbot POC case.

A senior developer's 24-hour deep test of Grok 4.5: a 1.5T-param MoE model at $2/M input tokens, with coding benchmarks rivaling GPT-5.5. Real performance, token efficiency, and limits explained.

Open weight ≠ runnable locally. This article breaks down the hardware barriers, VRAM limits, electricity costs, and parallelism constraints of models like GLM 5.2 and DeepSeek — revealing where open-weight models truly add value: driving cloud competition, not home replication.

This week in AI: OpenAI launches GPT-5.6 in three tiers (Sol/Terra/Luna) hitting 91.9% on coding benchmarks; DeepSeek and PKU open-source DSpark for 85% faster inference; Prime Intellect trains trillion-param models on just 28 H200s; Anthropic Claude enters Slack.

Unsloth v0.1.463-beta fixes a Studio crash caused by access-denied errors during llama-server service discovery. Improves stability for multi-user servers and Windows environments.

A systematic guide to the three cores of OpenAI LLM app development: GPT-4/GPT-3.5 model selection, token billing and cost-saving tips, and practical use of the Models, Completion, and Chat Completion APIs.

Want to learn Python from scratch but don't know where to begin? This article breaks down three stages—basic syntax, advanced mastery, and hands-on practice—with real projects in crawling, automation, and data analysis to help you build programming thinking.

Google's packed AI week: Gemini 3.5 Live Translate, NotebookLM agentic upgrades, DiffusionGemma text diffusion, and Project Genie. A full breakdown of Google's latest AI moves.

GPT-5.6 launches Soul/Terra/Luna, with flagship Soul scoring 91.9% on Terminal Bench 2.1. This article breaks down the Ultra vs Max reasoning modes, three-tier pricing, and four hidden pitfalls to guide your technical selection.

NVIDIA introduces Nonuniform Tensor Parallelism, letting GPUs bear different compute loads so training can continue without checkpoint rollback during hardware failures—boosting LLM training Goodput and fault tolerance at scale.