ki-modelle Claude
Claude Sonnet 5.5 Release September 2026 - All the Info
Claude Sonnet 5.5: Anthropic's mid tier next to Opus 5.5. Benchmarks, effort-based cost, breaking changes and how to choose vs Opus 5.5 and GPT-6.1 Sol.
Redaktion
Encyclopedia
boostN app encyclopedia: in-depth explanations of AI and app concepts.
-
- ki-modelle GPT
GPT-6.1 Sol Release September 2026 - All the Info
GPT-6.1 Sol: near Astra performance at a fifth of the cost. Benchmarks, pricing, limits and a selection guide against Opus 5.5 and Sonnet 5.5.
Redaktion - ki-modelle Claude
Claude Fable: Anthropic's Top Reasoning Line
Claude Fable is Anthropic's top reasoning line — above Opus, pricier, slower. Positioning, where it fits and how to choose it.
Redaktion - ki-modelle Claude
Claude Haiku: Anthropic's Fast, Low-Cost Line
Claude Haiku is Anthropic's fast, low-cost line for high volume. Positioning, where it fits and how to choose it.
Redaktion - ki-modelle Claude
Claude Opus: Anthropic's Pro Tier Explained
Claude Opus is Anthropic's pro tier for coding, agents and knowledge work — what the line stands for, where it fits and how to pick it.
Redaktion - ki-modelle Claude
Claude Sonnet: Anthropic's Balanced Line
Claude Sonnet is Anthropic's balanced line: more capable than Haiku, cheaper than Opus. Positioning, where it fits and how to choose it.
Redaktion - ki-modelle Mistral
Codestral — Mistral's code model line explained
Codestral is Mistral AI's code specialist: autocomplete, function generation, fill-in-the-middle. What the line is for, how it differs, when to pick it.
Redaktion - ki-modelle DeepSeek
DeepSeek V — the open model line at a glance
DeepSeek V as an open model line: positioning, use profile and how to choose. Flagship DeepSeek V4-Pro — cheap and self-hostable.
Redaktion - ki-modelle Gemini
Gemini Flash — Google's fast model line
Gemini Flash is Google's fast, low-cost model line: high throughput and native multimodality. Positioning, use profile and a selection guide.
Redaktion - ki-modelle Gemini
Gemini Pro — Google's reasoning spearhead
Gemini Pro is Google's reasoning spearhead: the deepest tier of the Gemini family for complex work. Positioning, use profile and a selection guide.
Redaktion - ki-modelle Zhipu AI
GLM (Zhipu AI / Z.ai) — the open model line at a glance
GLM by Zhipu AI as an open line: positioning, use profile and how to choose. Flagship GLM-5.3 with strong coding performance.
Redaktion - ki-modelle GPT
GPT Luna — the fast GPT-5.6 entry tier
GPT Luna is the entry tier of the GPT-5.6 family: fast and cheap for high volume. Positioning, use profile and a selection guide in one lexicon entry.
Redaktion - ki-modelle GPT
GPT Sol — OpenAI's top-tier line
GPT Sol is OpenAI's flagship line in the GPT-5.6 family: the deepest reasoning and coding. Positioning, use profile and a selection guide in one entry.
Redaktion - ki-modelle GPT
GPT Terra — the balanced GPT-5.6 tier
GPT Terra is the middle tier of the GPT-5.6 family: balanced across capability, speed and price. Positioning, use profile and a selection guide.
Redaktion - ki-modelle Grok
Grok (xAI) — the model line at a glance
Grok by xAI as a model line: positioning, use profile and how to choose. Current flagship Grok 4.6 with real-time access to X.
Redaktion - ki-modelle Moonshot AI
Kimi (Moonshot AI) — the open model line at a glance
Kimi by Moonshot AI as an open line: positioning, use profile and how to choose. Flagship Kimi K3 with a very large context window.
Redaktion - ki-modelle Mistral
Magistral — Mistral's reasoning line explained
Magistral is Mistral AI's reasoning line: a transparent chain of thought, strong in European languages, partly open weights. When to pick it.
Redaktion - ki-modelle Qwen
Qwen (Alibaba) — the model line at a glance
Qwen by Alibaba as a line: open dense/MoE models plus a proprietary Max line. Flagship Qwen 3.8-Max — positioning and how to choose.
Redaktion - ki-modelle DeepSeek
DeepSeek V4.1-Flash Release September 2026 - All the Info
DeepSeek V4.1-Flash: rank 1 on coding under 1h, sharp drop on agentic work over 1h. Prices, benchmarks and an honest read in the lexicon.
Redaktion -
Daybreak Blue: Security access to GPT-5.6 Sol explained
Daybreak Blue is not its own AI model, but OpenAI's vetted access to Sol with fewer security refusals for authorized defenders.
Redaktion - ki-orchestrierung Multi-Agent
A2A (Agent2Agent Protocol) — Communication Standard for Multi-Agent Systems
A2A is an open protocol for communication between independent AI agents across vendors — how it differs from MCP, and use cases in multi-agent systems.
Redaktion -
Agent Evaluation — Task Success Rate, Eval Harnesses, Benchmarks
How to evaluate AI agents: eval harnesses, task success rate, benchmarks like GAIA and Tau-Bench, LLM-as-a-judge — and how it differs from pure model quality.
Redaktion - ki-konzepte Agenten
Agent Memory — Short-Term and Long-Term Memory for AI Agents
Short- vs. long-term memory in AI agents: context window, external memory stores and vector DBs — how agents keep context across sessions.
Redaktion - ki-orchestrierung Workflow-Automatisierung
Agent Observability and Tracing — Spans, Logs and Debugging AI Agents
How tracing makes agent steps and tool calls visible as spans — the basis for debugging and cost control, with tools like LangSmith and Langfuse.
Redaktion - ki-orchestrierung Workflow-Automatisierung
State Management and Checkpointing
How agent workflows save state after every step (checkpointing) and resume exactly where they stopped after a crash — durable execution explained.
Redaktion - ki-konzepte API-Nutzung
Structured Outputs
How Structured Outputs at Anthropic, OpenAI and Gemini enforce JSON schemas via constrained decoding — validated output for pipelines, not free text.
Redaktion - arbeiten-mit-ki Content-Erstellung
Watermarks in AI Text: What Actually Gets Marked
How AI providers invisibly mark text, what Anthropic has done since August 2026, where the methods fail, and what it means for publishers.
Redaktion -
GPT-6 Astra Release September 2026 - All the Info
GPT-6 Astra is here: released September 2026. All the info on computer use, the Critical risk tier, reasoning, strengths & limits — in one lexicon entry.
Redaktion -
Agentic Workflows vs. Agents — when to fix, when to free
The difference between fixed, predefined workflows and autonomous AI agents — with Anthropic's definition, the trade-offs, and a clear decision guide.
Redaktion - ki-orchestrierung Workflow-Automatisierung
Evaluator-Optimizer Loop
One AI agent generates, a second one evaluates and critiques — looping around until the result meets a clear quality bar.
Redaktion -
Function Calling / Tool Use
How an LLM uses tools: define a tool as a schema, the model picks the function and arguments, the result returns to the chat — the basis of every agent.
Redaktion - ki-orchestrierung Workflow-Automatisierung
Human-in-the-Loop — Approval Gates for AI Agents
What human-in-the-loop means in agent workflows: approval gates, intervention points before critical actions, and why they are mandatory for irreversible steps.
Redaktion - ki-orchestrierung Keyword-Routing
LLM Routing and the LLM Router
What an LLM router does: send requests to cheap or strong models automatically, cut costs, and understand the risks when it misclassifies.
Redaktion - ki-orchestrierung MCP-Workflows
Model Context Protocol (MCP)
MCP is the open standard that connects AI models to external tools and data sources. Here is how its client-server architecture works.
Redaktion - ki-orchestrierung Multi-Agent
Multi-Agent Orchestration
How multiple AI agents work together: orchestrator-worker, supervisor pattern, task splitting and synthesis, and when multi-agent actually pays off.
Redaktion - ki-orchestrierung Agenten
Plan-and-Execute — the Planner-Executor Pattern for AI Agents
Plan-and-Execute means: build the full plan first, then work through it step by step. How the pattern works and where it beats ReAct on cost and quality.
Redaktion - ki-orchestrierung Workflow-Automatisierung
Prompt Chaining
Chaining several LLM calls into a pipeline: one step's output becomes the next step's input, gates as checks, and when chaining beats a mega-prompt.
Redaktion - ki-orchestrierung Agenten
The ReAct Pattern: Reasoning and Acting in Turns
How the ReAct pattern's Thought, Action and Observation loop works, why tool-using AI agents rely on it — and where it typically breaks down.
Redaktion - ki-werkzeuge Agenten-Frameworks
Agent Frameworks at a Glance
CrewAI, AutoGen, AutoGPT, DSPy, Browser Use, LangGraph — what agent frameworks do, where they differ, and when to reach for which one.
Redaktion - arbeiten-mit-ki Content-Erstellung
Content Creation with AI: a Clean Workflow
Just generating isn’t enough. The reliable flow: briefing, draft, fact-check, voice, SEO, human final edit — and where the human stays mandatory.
Redaktion - ki-konzepte LLM-Grundlagen
Embeddings Simply Explained
An embedding is a vector of numbers that places meaning in space. Why similar content sits close together and what embeddings are used for.
Redaktion - ki-modelle Open-Weight
FLUX.2
FLUX.2 by Black Forest Labs (Freiburg): open-weight image model, variants from 4B to 32B, licenses, hardware needs and how it stacks up in the market.
Redaktion - ki-werkzeuge Coding-CLI
Google Antigravity
Google's agent-first dev suite of desktop app, CLI and SDK. How Antigravity works, what it replaces from Gemini CLI, and how it competes with Claude Code.
Redaktion - ki-konzepte LLM-Grundlagen
LLM Hallucinations — Causes and Remedies
Why LLMs confidently invent falsehoods, what types of hallucinations exist, and which remedies actually help — RAG, source enforcement, verification.
Redaktion - ki-werkzeuge Hugging Face
The Hugging Face Ecosystem
Hub, Spaces and the Transformers, Datasets, Diffusers and Accelerate libraries — how they fit together and how the path to deployment works.
Redaktion - arbeiten-mit-ki Workflows
Automating AI Workflows
No-/low-code automation with Make, Zapier and n8n: building blocks, use cases for agencies and SMBs, pricing models and the costliest pitfalls.
Redaktion - ki-konzepte LLM-Grundlagen
The Context Window of LLMs — What Fits and What Doesn't
What the context window is, how big modern windows are, the lost-in-the-middle phenomenon, cost and latency, and the distinction from RAG and long-term memory.
Redaktion - ki-werkzeuge Agenten-Frameworks
LangGraph
LangGraph as the standard for agent orchestration — nodes, edges, state, loops, human-in-the-loop and persistence explained clearly.
Redaktion - ki-werkzeuge Agenten-Frameworks
Microsoft Agent Framework
Microsoft's SDK for AI agents: how it merges Semantic Kernel and AutoGen, what separates agents from workflows, and when it's worth adopting.
Redaktion - ki-werkzeuge Agenten-Frameworks
Microsoft Project Polaris
Microsoft's own AI coding model, unveiled at Build 2026. It replaces GPT-4 in GitHub Copilot from August 2026 — here is what it is.
Redaktion - ki-modelle Gemini
Nano Banana
Google's image model nicknamed Nano Banana — what hides behind the codename, what it can do, and how it differs from FLUX.2 and other image models.
Redaktion - arbeiten-mit-ki Produktivität
Working Productively in Parallel with AI Agents
The orchestrator-worker pattern, when parallel agents pay off and when not, the mechanics of shared queues, and the token price for it.
Redaktion - ki-werkzeuge Workflow-Automation
Self-Hosted Agent Stack (n8n, Dify, Ollama)
n8n, Dify and Ollama as a self-hosted AI stack — who does which layer, when it pays off and what hardware you actually need.
Redaktion - ki-konzepte LLM-Grundlagen
Temperature, Top-p and Sampling in LLMs
How an LLM picks the next token: temperature sharpens or flattens the probabilities, top-p and top-k limit the choice. Which setting for what.
Redaktion - arbeiten-mit-ki Token-Budget
Managing Your Token Budget — the Biggest Cost Levers for LLMs
Why tokens equal cost, why output is pricier than input, and the most effective levers: prompt caching, batch, lean context, model choice, output limit.
Redaktion - ki-konzepte LLM-Grundlagen
Tokens and Tokenization in LLMs Explained
What a token is, how tokenizers split text, and why tokens drive cost, context window, and speed — with rules of thumb for estimating token counts.
Redaktion - ki-werkzeuge Vektor-Datenbanken
Vector Databases Compared
What vector databases do, when you need one, and how Chroma, Weaviate, Milvus, Qdrant and pgvector stack up against each other.
Redaktion -
Comparing AI models — who builds what and how to choose
The major model families in 2026 at a glance. Who builds Claude, GPT, Gemini, Llama, Mistral, DeepSeek, Qwen — and which model to pick when.
Redaktion -
LLM Model Families at a Glance
Claude, GPT, Gemini, Llama & co. — who builds what, where each family shines, and how to pick the right model for your own use case.
Redaktion - ki-konzepte RAG
RAG Tuning and Balancing — Embeddings, Weights, Decay and MCP for Advanced Setups
How to steer retrieval on purpose: embedding choice, hybrid weights, reranker cascades, time decay, authority boost, MMR and MCP as a retrieval tool.
Redaktion - ki-konzepte Agenten
Building AI Agents — From Tool Use to Multi-Agent Systems
How AI agents work: from a single tool call through MCP, structured outputs and LangGraph to the question of when multi-agent setups actually pay off.
Redaktion - ki-konzepte API-Nutzung
Working with the LLM API — Streaming, Caching, Rate Limits
Practical API mechanics beyond pricing: streaming for UX, prompt caching against token cost, the Batch API for bulk jobs, rate limits without 429 drama.
Redaktion - ki-werkzeuge Pipeline-Frameworks
Pipeline Frameworks — LangChain, LlamaIndex and when you actually need them
LangChain, LlamaIndex, LangGraph and Haystack compared. What they're built for, when rolling your own pays off — and the criticisms worth taking seriously.
Redaktion -
AI Coding Tools Compared
Cursor, Windsurf, Claude Code, GitHub Copilot, Continue.dev, Aider compared. With table and decision guide for four typical developer workflows.
Redaktion - ki-konzepte Evaluation
Measuring LLM Quality — Evals, Benchmarks, Judges
How to tell whether an LLM system actually works: three evaluation layers from benchmarks to CI evals, plus pitfalls like Goodhart and judge bias.
Redaktion - ki-modelle LLM-Pricing
AI Pricing Made Simple
How AI models bill — tokens, input vs. output, hidden cost drivers and three levers to save. With price table and worked examples.
Redaktion - ki-konzepte Training & Fine-Tuning
Fine-Tuning Explained — When, How, With What
When fine-tuning pays off, which methods exist (SFT, DPO, LoRA, QLoRA), and what AMD vs. NVIDIA hardware actually means in practice.
Redaktion -
Running LLMs locally — hardware, tools, models
How to run language models on your own hardware — VRAM requirements, tooling (Ollama, LM Studio, llama.cpp, vLLM) and which models fit which GPU.
Redaktion - ki-konzepte Prompting
Prompt Security — Injection, Leaking, Guardrails
How prompt injection, prompt leaking, and jailbreaks work — and which defenses (guardrails, spotlighting, sanitization) actually help.
Redaktion - ki-konzepte Prompting
Prompting Techniques — An Overview
Zero-shot, few-shot, chain-of-thought, tree of thoughts, ReAct & co. — when each prompting technique pays off and how they fit together.
Redaktion - ki-konzepte RAG
Understanding RAG — From Idea to Production System
How a RAG pipeline works: embedding, vector DB, retrieval, reranking, prompt — and which pitfalls show up in practice.
Redaktion - ki-modelle Muse
Meta Muse Spark April 2026 - All the Info
Muse Spark is Metas first frontier model from Superintelligence Labs: closed-weight, thought compression, multimodal. Facts, benchmarks and where it fits.
Redaktion