$/MTok (Cost per Million Tokens)
Standard pricing unit for AI APIs — cost in US dollars per one million processed tokens. Listed separately for input, output and sometimes cache.
Redaktion
KI-Konzepte
Grundbegriffe und Mechanik moderner Sprachmodelle und KI-Workflows.
-
- ki-konzepte Training & Fine-Tuning
Adapter (PEFT)
An adapter is a small, freshly added network module injected into a frozen language model — only these modules are trained, while the base model itself stays unchanged.
Redaktion - ki-konzepte RAG
Agentic RAG
Agentic RAG is a RAG architecture in which an agent drives retrieval — it decomposes the question, searches iteratively, checks the hits, and decides itself whether a second or third pass is needed.
Redaktion - ki-konzepte Agenten
AI Agent
An AI agent combines a language model with tools and works toward a goal across multiple steps — typically in a loop of observing, planning and acting.
Redaktion -
Attention Mechanism
Computational method in transformer models that weights which parts of the input text are most relevant for predicting the next token — instead of treating every part as equally important.
Redaktion - ki-konzepte LLM-Grundlagen
Audio normalization
Audio normalization is the adaptive linear scaling of an audio signal to a target level — the tool first measures a reference value (peak, RMS or LUFS) and from that calculates the required gain.
Redaktion - ki-konzepte Agenten
AutoGPT
AutoGPT is an open-source experiment released in 2023 that runs a language model such as GPT-4 in an autonomous loop — it breaks a goal into subtasks itself, calls tools and works through them with minimal human intervention.
Redaktion - ki-konzepte LLM-Grundlagen
Automatic Speech Recognition (ASR)
Automatic Speech Recognition (ASR), also called Speech-to-Text (STT), is the automatic conversion of spoken language into written text by an acoustic model.
Redaktion -
Batch API
Asynchronous API mode that collects many requests and processes them at a significant discount — results are typically delivered within 24 hours.
Redaktion - ki-konzepte Lokales LLM
Batch Inference
Batch inference is the bundled processing of many prompts in one pass — locally on a GPU or as an asynchronous API job — trading real-time latency for throughput and cost.
Redaktion - ki-konzepte Evaluation
Benchmark (AI)
An AI benchmark is a standardized task suite that makes language models comparable — same prompts, same scoring, transparent results for reasoning, knowledge, or coding.
Redaktion - ki-konzepte Agenten
Browser Use
Browser use refers to AI agents that drive a real web browser — navigating, clicking, filling forms, reading content — to complete web tasks autonomously. The browser serves as both the tool and the perception surface.
Redaktion - ki-konzepte Training & Fine-Tuning
Catastrophic Forgetting
Catastrophic forgetting describes the effect that a neural network loses previously acquired knowledge — wholly or partially — when it is trained further on new data; a key problem when fine-tuning language models.
Redaktion - ki-konzepte Prompting
Chain-of-Thought
Chain-of-Thought (CoT) is a prompting technique that asks the model to spell out its reasoning in intermediate steps — boosting accuracy on multi-step tasks.
Redaktion - ki-konzepte RAG
Chunking
Chunking is the splitting of longer texts into smaller, self-contained pieces — the first step of any RAG pipeline because embeddings and retrieval work at the chunk level.
Redaktion - ki-konzepte Agenten
Computer Use
Computer Use is a capability of modern AI models that lets them operate a computer like a human — see the screen, move the mouse, use the keyboard — to perform tasks across arbitrary applications without an API.
Redaktion - ki-konzepte Prompting
Constitutional AI
Constitutional AI is a training method developed by Anthropic in which an AI model critiques and revises its own behaviour against an explicit list of principles (a "constitution") — rather than relying solely on human ratings.
Redaktion - ki-konzepte Prompting
Context Engineering
Context engineering is the discipline of deliberately curating and maintaining everything inside the LLM context window — system prompt, tool definitions, RAG hits, memory, conversation history.
Redaktion - ki-konzepte Evaluation
Context Precision
Context Precision is a RAG evaluation metric (0–1) measuring how well the retriever ranks relevant contexts above irrelevant ones — the focus is ranking quality, not raw hit count.
Redaktion - ki-konzepte Evaluation
Context Recall
Context Recall is a RAG evaluation metric (0–1) measuring how completely the retriever supplied the facts needed for the gold answer — every missing fact lowers the score.
Redaktion - ki-konzepte LLM-Grundlagen
Context window
The context window is the maximum number of tokens a language model can process at once — input and output combined.
Redaktion - ki-konzepte RAG
Contextual Retrieval
Contextual Retrieval is an Anthropic RAG technique that prepends an LLM-generated context sentence to every chunk before indexing — sharply raising retrieval quality.
Redaktion - ki-konzepte Training & Fine-Tuning
Continued Pretraining
Continued pretraining is the further training of an already pretrained language model on large amounts of domain text — before any classic fine-tuning begins.
Redaktion - ki-konzepte RAG
Cosine Similarity
Cosine similarity measures how similar two vectors are by the angle between them — the default metric in RAG retrieval for comparing query embeddings against the embeddings stored in a vector index.
Redaktion - ki-konzepte LLM-Grundlagen
dBFS, Headroom & Clipping
Three related concepts for digital audio levels: dBFS is the scale with 0 dBFS as the hard ceiling, headroom is the safety margin below the maximum, clipping is the hard cut-off when a signal exceeds it.
Redaktion - ki-konzepte Training & Fine-Tuning
DPO (Direct Preference Optimization)
DPO is a fine-tuning method that aligns language models directly to human preference pairs — without a separate reward model and without reinforcement learning.
Redaktion - ki-konzepte LLM-Grundlagen
Embedding
An embedding is a numeric representation (vector) of text, an image, or other data in which semantically similar content sits close together in space.
Redaktion - ki-konzepte Agenten
Ensemble / Multi-Model Orchestration
Ensemble means combining several deliberately varied LLM runs or models whose findings complement each other. Multi-model orchestration drives these runs via orchestrators with sub-agents, so the union of results is larger than any single run.
Redaktion - ki-konzepte Evaluation
Evals
Evals are systematic tests for LLM applications — fixed test cases, automatic scoring of the responses, result as a score. The foundation for objectively comparing prompt or model changes.
Redaktion - ki-konzepte Evaluation
Faithfulness
Faithfulness is a RAG evaluation metric (0–1) measuring how many claims in a response can actually be derived from the retrieved contexts — a direct hallucination indicator.
Redaktion - ki-konzepte Prompting
Few-Shot Prompting
Few-shot prompting is a technique that includes a handful of input/output examples in the prompt so the model picks up the desired format and style.
Redaktion - ki-konzepte Training & Fine-Tuning
Fine-Tuning
Continued training of a pretrained model on a smaller, specialized dataset to adapt style, format or domain knowledge in a targeted way.
Redaktion - ki-konzepte Agenten
Function Calling
Function calling is a language model's ability to produce a structured function invocation instead of a text reply — the technical foundation for tool use and AI agents.
Redaktion - ki-konzepte Lokales LLM
GGUF
GGUF (GPT-Generated Unified Format) is the standard file format for quantised LLMs on the llama.cpp engine — a single `.gguf` file holds weights, tokenizer and metadata together.
Redaktion - ki-konzepte RAG
GraphRAG
GraphRAG is a RAG variant that uses a knowledge graph in addition to vector search — answers come from linked entities and relationship paths, not just similar text.
Redaktion - ki-konzepte Prompting
Guardrails
Guardrails are protective layers around a language model that inspect inputs and outputs to catch unwanted behaviour — off-topic answers, PII leaks, unsafe actions.
Redaktion - ki-konzepte LLM-Grundlagen
Hallucination
A hallucination is a language-model response that sounds plausible but is factually wrong or fabricated — typically caused by statistical guessing without a factual anchor.
Redaktion - ki-konzepte RAG
Hybrid Search
Hybrid search combines lexical full-text search (e.g. BM25) with semantic vector search and unites the strengths of both — precise on keywords and robust on meaning.
Redaktion - ki-konzepte RAG
HyDE
HyDE (Hypothetical Document Embeddings) is a RAG technique where an LLM first drafts a fake answer to a query — and then uses that answer's embedding to find real documents.
Redaktion - ki-konzepte Prompting
Indirect Prompt Injection
Indirect prompt injection is an attack that hides malicious instructions inside external content — web pages, documents, emails — that an LLM then processes and unintentionally executes.
Redaktion - ki-konzepte LLM-Grundlagen
Inference
Inference is the act of running an already-trained model to turn an input into a response — production use, not training.
Redaktion -
Input Token
Tokens you send to an AI model in an API call — your prompt, the context, attached documents. Billed separately from output tokens and usually much cheaper.
Redaktion - ki-konzepte Training & Fine-Tuning
Instruction Tuning
Instruction tuning is a variant of supervised fine-tuning in which a language model is trained on a broad mix of instruction-response pairs — so it reliably follows instructions across arbitrary tasks.
Redaktion - ki-konzepte Prompting
Jailbreak
A jailbreak deliberately bypasses a language model's safety and behaviour rules through cleverly crafted prompts, making the model produce content it would normally refuse.
Redaktion - ki-konzepte API-Nutzung
Keyword Glossary (STT)
A keyword glossary for Speech-to-Text is a short, curated list of project-specific terms, brand and product names handed to an STT model as a context hint before transcription — typically through Whisper's `initial_prompt`.
Redaktion - ki-konzepte Training & Fine-Tuning
Knowledge Distillation
Knowledge distillation is a training technique in which a small student model learns from the behaviour of a large teacher model — aiming for comparable quality at significantly lower resource cost.
Redaktion - ki-konzepte LLM-Grundlagen
KV Cache
The KV cache (key-value cache) stores the already computed key and value vectors of a language model's attention layers during text generation. This avoids recomputing the entire context for each new token and significantly speeds up inference.
Redaktion - ki-konzepte LLM-Grundlagen
LLM
LLM stands for Large Language Model — a neural network trained on large volumes of text to interpret and generate natural language.
Redaktion - ki-konzepte Evaluation
LLM-as-a-Judge
LLM-as-a-Judge is the practice of using a language model as an evaluator — it compares responses from other models or scores outputs against given criteria.
Redaktion - ki-konzepte Training & Fine-Tuning
LoRA
Low-Rank Adaptation — parameter-efficient fine-tuning that trains only small additional matrices instead of all model weights.
Redaktion - ki-konzepte Agenten
MCP
MCP — Model Context Protocol — is an open standard from Anthropic for connecting AI models to external data sources and tools through a unified interface.
Redaktion - ki-konzepte Prompting
Meta-Prompting
Meta-prompting is the technique of having an LLM write or improve a prompt for a task rather than formulating it yourself — the prompt becomes the output.
Redaktion - ki-konzepte LLM-Grundlagen
Model Card
A model card is a standardised datasheet for an AI model — documenting purpose, training data, performance, limitations and ethical considerations in one place.
Redaktion - ki-konzepte Training & Fine-Tuning
ORPO (Odds Ratio Preference Optimization)
ORPO is a training method that unifies supervised fine-tuning and preference optimisation in a single step — leaner than the classic SFT-plus-RLHF pipeline.
Redaktion -
Output Token
Tokens an AI model produces as its response. Billed separately and usually three to five times more expensive than input tokens because the model has to actively generate them.
Redaktion - ki-konzepte LLM-Grundlagen
Path Dependency (LLM Output)
Path dependency describes how, during LLM generation, each early token shapes everything that follows — the first verbalized finding steers the entire remaining analysis in one direction.
Redaktion - ki-konzepte Training & Fine-Tuning
PEFT (Parameter-Efficient Fine-Tuning)
PEFT bundles methods that adapt a pretrained LLM by training only a small fraction of its weights — typically under 1%, instead of all billions of parameters.
Redaktion - ki-konzepte Training & Fine-Tuning
Prefix Tuning
Prefix tuning is a PEFT method that trains a sequence of learned "virtual tokens" prepended to the input rather than any model weights — the base model itself stays frozen.
Redaktion -
Prompt Caching
Prompt caching is an API feature in which a provider stores recurring prompt prefixes — making subsequent requests cheaper and faster because the cached portion is not reprocessed.
Redaktion - ki-konzepte Prompting
Prompt engineering
Prompt engineering is the discipline of steering AI models through carefully crafted input prompts — closer to precise writing than to classic programming.
Redaktion - ki-konzepte Prompting
Prompt Injection
Prompt injection is an attack that hides instructions inside an LLM's input data to make the model ignore or subvert its original instructions.
Redaktion - ki-konzepte Prompting
Prompt Leaking
Prompt leaking is an attack that tricks an LLM into revealing its hidden system prompt or other confidential context contents — a special case of prompt injection.
Redaktion - ki-konzepte Prompting
Prompt Template
A prompt template is a reusable prompt skeleton with placeholders that get filled with concrete values at runtime — the basis for reproducible LLM calls in applications.
Redaktion - ki-konzepte Training & Fine-Tuning
QLoRA
Extension of LoRA that quantizes the base model to 4 bits — allowing even very large LLMs to be fine-tuned on a single GPU.
Redaktion - ki-konzepte Lokales LLM
Quantization
Quantization reduces the numeric precision of model weights — e.g. from 16-bit float to 4-bit integer — making language models small and fast enough for consumer hardware, with manageable quality loss.
Redaktion - ki-konzepte RAG
Query Expansion
Query expansion enriches the original query in a RAG pipeline with synonyms, reformulations or hypothetical answer texts in order to lift retrieval recall.
Redaktion - ki-konzepte LLM-Grundlagen
RAG
RAG (Retrieval-Augmented Generation) connects a language model to an external knowledge source — relevant passages are retrieved and passed to the model alongside the question.
Redaktion - ki-konzepte Evaluation
RAGAS
RAGAS is an open-source framework for automated evaluation of RAG and agent pipelines — with built-in metrics like faithfulness, context precision, and answer relevancy.
Redaktion -
Rate Limit (AI)
Provider-enforced cap on requests or tokens per time window — it protects infrastructure and ensures fair usage across customers.
Redaktion - ki-konzepte Prompting
ReAct (Prompting)
ReAct is a prompting pattern that has an LLM alternate between reasoning (thoughts) and action (tool calls) — the foundation behind many agent implementations.
Redaktion - ki-konzepte LLM-Grundlagen
Reasoning Effort
Reasoning Effort is a control parameter on modern reasoning models that sets how much internal step-by-step thinking (thinking tokens) a model spends before answering — higher levels raise quality but also latency and cost.
Redaktion - ki-konzepte Evaluation
Recall
Recall (hit rate, sensitivity) measures the share of all actually existing relevant cases that a system finds. Formula: relevant cases found divided by all relevant cases that truly exist. Requires a known ground truth.
Redaktion - ki-konzepte RAG
Reranking
Reranking reorders an already-retrieved list of hits with a more accurate model — typically a cross-encoder that scores query and each candidate together, instead of just comparing vector distances.
Redaktion - ki-konzepte Training & Fine-Tuning
RLHF
Reinforcement Learning from Human Feedback — training procedure that aligns models with helpful and safe behavior using human preference comparisons.
Redaktion - ki-konzepte Prompting
Role Prompting
Role prompting assigns the LLM a concrete role or persona ("you are an experienced tax lawyer …") to steer style, vocabulary and depth of answers.
Redaktion - ki-konzepte LLM-Grundlagen
Safetensors
Safetensors is a file format developed by Hugging Face for storing model weights — faster, safer and more language-agnostic than the older PyTorch `.pt`/`.bin` format.
Redaktion - ki-konzepte LLM-Grundlagen
Sampling (LLM)
Sampling is the weighted drawing of the next token from a language model's probability distribution — governed by temperature, top-p and top-k. It produces the variability between two runs.
Redaktion - ki-konzepte Prompting
Self-Consistency
Self-Consistency is a prompting technique where the same question is answered multiple times with chain-of-thought, and the most frequent answer is picked as the final result.
Redaktion - ki-konzepte Prompting
Self-Refine
Self-Refine is a prompting technique in which a model critiques its own output and improves it across several iterations — without human feedback.
Redaktion - ki-konzepte RAG
Semantic Search
Semantic search retrieves content by meaning rather than exact word match — query and documents are compared as embeddings, so synonyms and paraphrases also hit.
Redaktion - ki-konzepte Training & Fine-Tuning
SFT (Supervised Fine-Tuning)
SFT is the supervised post-training of a pretrained language model on a dataset of input-output pairs — the step that turns a base model into a usable assistant.
Redaktion - ki-konzepte Prompting
Spotlighting
Spotlighting is a defense technique against prompt injection in which untrusted inputs are marked so the model treats them as data — not as instructions.
Redaktion - ki-konzepte LLM-Grundlagen
Stop Sequences
Stop sequences are strings that make an LLM end token generation as soon as they appear. They constrain the output deliberately — to prevent role switches, format markers or overly long answers.
Redaktion - ki-konzepte API-Nutzung
Streaming (LLM)
Streaming means transmitting an LLM's response token by token in real time — the user sees the text word by word rather than only after the full generation completes.
Redaktion - ki-konzepte Agenten
Structured Output / JSON Mode
Structured output is a language model's ability to deliver responses in a defined schema (typically JSON), reliable enough to be consumed directly by downstream code.
Redaktion - ki-konzepte Training & Fine-Tuning
Synthetic Data
Synthetic data is artificially generated training or evaluation data — typically produced by a language model itself to extend datasets without needing real-world sources.
Redaktion - ki-konzepte Prompting
System Prompt
A system prompt is the instruction that shapes a language model's behavior across an entire conversation — in contrast to the user prompt, which varies per message.
Redaktion - ki-konzepte LLM-Grundlagen
Temperature
Temperature is a sampling parameter that controls how deterministic or creative a language model's responses are — low values stay conservative, high values produce more variety.
Redaktion - ki-konzepte LLM-Grundlagen
Thinking Budget
The thinking budget (token budget) is the maximum number of internal reasoning tokens a reasoning model may spend on "thinking" per request — controlling the trade-off between answer depth and cost or latency.
Redaktion -
Tier Pricing
Tiered model line-up from a provider — small fast variants (Mini/Flash/Haiku) at a fraction of the price of the big frontier models. Also: volume tiers with quantity discounts.
Redaktion - ki-konzepte LLM-Grundlagen
Token
A token is the smallest unit a language model works with internally — usually a sub-word fragment, occasionally a single character.
Redaktion - ki-konzepte LLM-Grundlagen
Tokenizer
A tokenizer is the program that splits text into tokens before a language model can process it — it determines how many tokens a piece of text costs.
Redaktion - ki-konzepte Agenten
Tool Call
An invocation of a tool by an AI model during a conversation — such as reading a file, running a bash command, fetching from the web or calling an MCP tool. The foundation of agentic workflows.
Redaktion - ki-konzepte LLM-Grundlagen
Top-p / Top-k
Top-p (nucleus sampling) and top-k are sampling strategies for LLMs that decide which token candidates to draw from at each step — together with temperature they control the creativity of the output.
Redaktion - ki-konzepte API-Nutzung
TPM/RPM
TPM (tokens per minute) and RPM (requests per minute) are the two common units in which AI API providers express their rate limits — TPM caps the token volume, RPM caps the number of requests per minute.
Redaktion - ki-konzepte Prompting
Tree of Thoughts
Tree of Thoughts (ToT) is a prompting technique in which an LLM explores multiple solution paths as a branching tree, scores them and follows only the promising branches — a generalisation of Chain-of-Thought.
Redaktion - ki-konzepte RAG
Vector Database
A vector database stores embeddings as high-dimensional vectors and searches them by semantic similarity — the core infrastructure for RAG, semantic search, and recommendation systems.
Redaktion - ki-konzepte Lokales LLM
VRAM
VRAM (Video RAM) is the dedicated memory on a GPU — the single most important hardware figure for running language models locally, because ideally the whole model and its context must fit inside it.
Redaktion - ki-konzepte LLM-Grundlagen
Whisper
Whisper is OpenAI's open speech-to-text model from 2022 — a multilingual encoder-decoder transformer that turns audio into text and ships in several sizes under an MIT license.
Redaktion - ki-konzepte API-Nutzung
Whisper initial_prompt
The `initial_prompt` parameter lets you pass Whisper a keyword list or sample sentence before transcription. The model treats this text as context and recognises the included terms far more reliably.
Redaktion - ki-konzepte Prompting
XML tags in the prompt
XML tags are named brackets like <context>...</context> used to clearly separate sections of a prompt. They structure the input, split instruction from data and make the model's output more predictable.
Redaktion - ki-konzepte Prompting
Zero-Shot
Zero-shot describes solving a task with a language model without providing any examples — purely via the task description in the prompt.
Redaktion