LLM Wiki

<h1>LLM Wiki Index</h1> <p>Welcome to the LLM Wiki — a comprehensive reference covering Large Language Models, their architectures, training methods, inference techniques, ecosystem, and the AI Empire blog operations. Built using the <a href="/wiki/wiki-schema" class="wiki-link-dead">SCHEMA</a> conventions.</p> <hr> <h2>LLM Domain</h2> <h3>Entities</h3> <h4>Frontier Models</h4> <ul> <li><a href="/wiki/entities/gpt-4o">GPT-4o</a> — OpenAI&#39;s flagship multimodal model (May 2024)</li> <li><a href="/wiki/gpt-4o-mini" class="wiki-link-dead">gpt-4o-mini</a> — OpenAI&#39;s small, fast, cheap model</li> <li><a href="/wiki/o1" class="wiki-link-dead">o1</a> — OpenAI&#39;s reasoning model (Sept 2024)</li> <li><a href="/wiki/entities/claude-4">Claude 4</a> — Anthropic&#39;s Claude 4 family (Opus, Sonnet)</li> <li><a href="/wiki/entities/gemini-2-pro">Gemini 2 Pro</a> — Google&#39;s Gemini 2 Pro model</li> <li><a href="/wiki/gemini-2-flash" class="wiki-link-dead">gemini-2-flash</a> — Google&#39;s Gemini 2 Flash for speed</li> <li><a href="/wiki/entities/llama-4">Llama 4</a> — Meta&#39;s Llama 4 family</li> <li><a href="/wiki/entities/deepseek-v3">DeepSeek V3</a> — DeepSeek&#39;s V3 dense MoE model</li> <li><a href="/wiki/entities/deepseek-r1">DeepSeek R1</a> — DeepSeek&#39;s first reasoning model</li> <li><a href="/wiki/entities/mistral-large">Mistral Large</a> — Mistral AI&#39;s flagship large model</li> <li><a href="/wiki/qwen-2.5" class="wiki-link-dead">qwen-2.5</a> — Alibaba&#39;s Qwen 2.5 family</li> <li><a href="/wiki/qwen-2.5-coder" class="wiki-link-dead">qwen-2.5-coder</a> — Qwen specialized for code</li> </ul> <h4>Mid-Size &amp; Open Models</h4> <ul> <li><a href="/wiki/llama-3.2" class="wiki-link-dead">llama-3.2</a> — Meta&#39;s Llama 3.2 series (1B-90B)</li> <li><a href="/wiki/mistral-small" class="wiki-link-dead">mistral-small</a> — Mistral AI&#39;s efficient small model</li> <li><a href="/wiki/phi-3" class="wiki-link-dead">phi-3</a> — Microsoft&#39;s small language model</li> <li><a href="/wiki/gemma" class="wiki-link-dead">gemma</a> — Google&#39;s open model family</li> </ul> <h4>Companies</h4> <ul> <li><a href="/wiki/openai" class="wiki-link-dead">openai</a> — Creator of GPT series, o1, DALL-E</li> <li><a href="/wiki/anthropic" class="wiki-link-dead">anthropic</a> — Creator of Claude series</li> <li><a href="/wiki/google-deepmind" class="wiki-link-dead">google-deepmind</a> — Creator of Gemini series</li> <li><a href="/wiki/meta-ai" class="wiki-link-dead">meta-ai</a> — Creator of Llama series</li> <li><a href="/wiki/mistral-ai" class="wiki-link-dead">mistral-ai</a> — French AI lab, Mistral model series</li> <li><a href="/wiki/deepseek" class="wiki-link-dead">deepseek</a> — Chinese AI lab, DeepSeek series</li> <li><a href="/wiki/alibaba-qwen" class="wiki-link-dead">alibaba-qwen</a> — Alibaba&#39;s Qwen model team</li> </ul> <h3>Concepts</h3> <h4>Architecture</h4> <ul> <li><a href="/wiki/concepts/transformer-architecture">Transformer Architecture</a> — The foundational architecture for all modern LLMs</li> <li><a href="/wiki/concepts/attention-mechanism">Attention Mechanism</a> — How models focus on relevant parts of input</li> <li><a href="/wiki/multi-head-attention" class="wiki-link-dead">multi-head-attention</a> — Parallel attention computation</li> <li><a href="/wiki/concepts/mixture-of-experts">Mixture of Experts</a> — Sparse activation architecture</li> <li><a href="/wiki/concepts/kv-cache">KV Cache</a> — Key-value caching for efficient inference</li> <li><a href="/wiki/concepts/context-window">Context Window</a> — Maximum sequence length a model can process</li> <li><a href="/wiki/positional-encoding" class="wiki-link-dead">positional-encoding</a> — How models understand token order</li> <li><a href="/wiki/rope" class="wiki-link-dead">rope</a> — Rotary Position Embeddings</li> <li><a href="/wiki/grouped-query-attention" class="wiki-link-dead">grouped-query-attention</a> — GQA for efficient attention</li> </ul> <h4>Training</h4> <ul> <li><a href="/wiki/pretraining" class="wiki-link-dead">pretraining</a> — Initial large-scale training on unlabeled data</li> <li><a href="/wiki/supervised-fine-tuning" class="wiki-link-dead">supervised-fine-tuning</a> — Task-specific training on labeled data</li> <li><a href="/wiki/concepts/rlhf">Reinforcement Learning from Human Feedback</a> — Reinforcement Learning from Human Feedback</li> <li><a href="/wiki/concepts/dpo">Direct Preference Optimization</a> — Direct Preference Optimization</li> <li><a href="/wiki/lora" class="wiki-link-dead">lora</a> — Low-Rank Adaptation for efficient fine-tuning</li> <li><a href="/wiki/quantization" class="wiki-link-dead">quantization</a> — Reducing model precision for efficiency</li> <li><a href="/wiki/distillation" class="wiki-link-dead">distillation</a> — Training smaller models from larger ones</li> <li><a href="/wiki/data-mixing" class="wiki-link-dead">data-mixing</a> — Strategies for blending training data sources</li> </ul> <h4>Inference</h4> <ul> <li><a href="/wiki/temperature" class="wiki-link-dead">temperature</a> — Controlling output randomness</li> <li><a href="/wiki/top-p-sampling" class="wiki-link-dead">top-p-sampling</a> — Nucleus sampling strategy</li> <li><a href="/wiki/top-k-sampling" class="wiki-link-dead">top-k-sampling</a> — Top-k token selection</li> <li><a href="/wiki/beam-search" class="wiki-link-dead">beam-search</a> — Beam search decoding</li> <li><a href="/wiki/streaming" class="wiki-link-dead">streaming</a> — Token-by-token output delivery</li> <li><a href="/wiki/structured-output" class="wiki-link-dead">structured-output</a> — JSON and schema-constrained generation</li> <li><a href="/wiki/function-calling" class="wiki-link-dead">function-calling</a> — Tool-use via model APIs</li> </ul> <h4>Optimization</h4> <ul> <li><a href="/wiki/prompt-engineering" class="wiki-link-dead">prompt-engineering</a> — Techniques for effective prompting</li> <li><a href="/wiki/chain-of-thought" class="wiki-link-dead">chain-of-thought</a> — Step-by-step reasoning prompting</li> <li><a href="/wiki/rag" class="wiki-link-dead">rag</a> — Retrieval-Augmented Generation</li> <li><a href="/wiki/few-shot-learning" class="wiki-link-dead">few-shot-learning</a> — In-context learning from examples</li> </ul> <h3>Comparisons</h3> <ul> <li><a href="/wiki/comparisons/open-vs-closed-models">Open-Source vs Closed-Source LLMs</a> — Comparing open-source vs closed-source LLMs</li> <li><a href="/wiki/api-vs-local" class="wiki-link-dead">api-vs-local</a> — API-based vs local model deployment</li> <li><a href="/wiki/comparisons/pricing-comparison">LLM API Pricing Comparison</a> — API pricing across providers</li> <li><a href="/wiki/context-window-comparison" class="wiki-link-dead">context-window-comparison</a> — Context window sizes across models</li> <li><a href="/wiki/coding-models-comparison" class="wiki-link-dead">coding-models-comparison</a> — Coding capability comparison</li> <li><a href="/wiki/reasoning-models-comparison" class="wiki-link-dead">reasoning-models-comparison</a> — Reasoning models comparison (o1, R1, Claude)</li> </ul> <hr> <h2>Blog Empire Domain</h2> <h3>Blog Entities</h3> <h4>Active Blogs (12)</h4> <ul> <li><a href="/wiki/entities/niteagent">NiteAgent</a> — AI agents, agentic workflows, production AI. Flagship blog with highest quality standards.</li> <li><a href="/wiki/entities/toolbrain">ToolBrain</a> — AI tools, frameworks, developer productivity. Highest volume (~298 posts).</li> <li><a href="/wiki/entities/codeintel">CodeIntel</a> — Code intelligence, PR automation, benchmark leaderboards.</li> <li><a href="/wiki/entities/nocodeinsider">NoCode Insider</a> — No-code platforms, low-code automation, workflow design.</li> <li><a href="/wiki/entities/shfg">Smart Home Field Guide</a> — Smart home technology, IoT, home automation. (Smart Home Field Guide)</li> <li><a href="/wiki/entities/hermestuts">Hermes Tutorials</a> — Hermes Agent tutorials, LLM Wiki, documentation.</li> <li><a href="/wiki/entities/quantbrainai">QuantBrainAI</a> — Quantitative finance, algorithmic trading, market analysis.</li> <li><a href="/wiki/entities/aigamingdev">AIGamingDev</a> — AI in game development, procedural generation, game AI.</li> <li><a href="/wiki/entities/cybersec-ai">CyberSec AI</a> — AI cybersecurity, threat detection, security automation. (New)</li> <li><a href="/wiki/entities/design-agent">Design Agent</a> — AI design tools, UI/UX automation, creative AI. (New)</li> <li><a href="/wiki/entities/healthywallet">Healthy Wallet</a> — Personal finance, budgeting, investing. (New)</li> <li><a href="/wiki/entities/careerml">CareerML</a> — ML engineering careers, AI job market, skill development.</li> <li><a href="/wiki/entities/stacksfree">StacksFree</a> — Free tech stacks, open-source tools, self-hosted infrastructure. (Internal)</li> </ul> <h4>Quality Status Overview</h4> <table> <thead> <tr> <th>Blog</th> <th>Status</th> <th>Citation Rate</th> <th>Posts</th> </tr> </thead> <tbody><tr> <td>niteagent</td> <td>✅ Pass</td> <td>99%</td> <td>~149</td> </tr> <tr> <td>codeintel</td> <td>✅ Pass</td> <td>85%</td> <td>~124</td> </tr> <tr> <td>shfg</td> <td>✅ Pass</td> <td>94%</td> <td>~32</td> </tr> <tr> <td>hermestuts</td> <td>✅ Pass</td> <td>67%</td> <td>~49</td> </tr> <tr> <td>toolbrain</td> <td>❌ Fail (unsourced)</td> <td>47%</td> <td>~298</td> </tr> <tr> <td>nocodeinsider</td> <td>❌ Fail (words + unsourced)</td> <td>88%</td> <td>~8</td> </tr> <tr> <td>quantbrainai</td> <td>❌ Fail (words + unsourced)</td> <td>52%</td> <td>~46</td> </tr> <tr> <td>aigamingdev</td> <td>❌ Fail (words + unsourced)</td> <td>72%</td> <td>~46</td> </tr> <tr> <td>careerml</td> <td>❌ Fail (unsourced)</td> <td>70%</td> <td>~20</td> </tr> <tr> <td>cybersec-ai</td> <td>❌ Fail (no posts)</td> <td>--</td> <td>0</td> </tr> <tr> <td>design-agent</td> <td>❌ Fail (no posts)</td> <td>--</td> <td>0</td> </tr> <tr> <td>healthywallet</td> <td>❌ Fail (no posts)</td> <td>--</td> <td>0</td> </tr> </tbody></table> <h3>Workflow Concepts</h3> <ul> <li><a href="/wiki/concepts/arena-workflow">Arena Workflow</a> — Subagent arena pipeline: model-vs-model comparison with builder/judge roles</li> <li><a href="/wiki/concepts/tool-building-workflow">Tool Building Workflow</a> — ToolBrain tool page pipeline: repo-prompt pattern</li> <li><a href="/wiki/concepts/content-pipeline">Content Pipeline</a> — End-to-end content generation: cron jobs, scheduling, 5-Gate quality chain</li> <li><a href="/wiki/concepts/leaderboard">Leaderboard</a> — Model performance tracking: benchmarks, builder vs judge roles, smart task routing</li> </ul> <h3>Comparisons (Blog Empire)</h3> <ul> <li><a href="/wiki/builder-vs-judge-models" class="wiki-link-dead">builder-vs-judge-models</a> — Builder vs judge model selection guide</li> <li><a href="/wiki/blog-quality-comparison" class="wiki-link-dead">blog-quality-comparison</a> — Quality metrics across all 12 blogs</li> </ul> <hr> <h2>Structure</h2> <ul> <li><strong>SCHEMA.md</strong> — <a href="/wiki/wiki-schema" class="wiki-link-dead">Full schema, conventions, and tag taxonomy (LLM + Blog Empire)</a></li> <li><strong>index.md</strong> — <a href="/wiki/llm-wiki-index" class="wiki-link-dead">This index page</a></li> <li><strong>log.md</strong> — <a href="/wiki/wiki-log" class="wiki-link-dead">Chronological change log</a></li> </ul> <h3>Raw Sources</h3> <ul> <li><code>raw/articles/</code> — Extracted web articles</li> <li><code>raw/papers/</code> — Extracted academic papers</li> </ul> <h3>Special</h3> <ul> <li><a href="/wiki/wiki-glossary" class="wiki-link-dead">wiki-glossary</a> — Glossary of terms</li> </ul>