Das Preisdesk

Den echten Preis der KI kennen, bevor Sie zahlen.

Ein unabhängiges Desk für aktuelle Preise, gerankte Vergleiche und die Suchen, die Käufer wirklich stellen. Veraltete Briefings verschwinden, statt auf der Startseite zu bleiben.

Die ausführlichen Briefings bleiben auf Englisch. Navigation und Entscheidungszentrale sind auf Deutsch.

Preise vergleichen Meine Kosten rechnen Empfehlung erhalten

New · GPT-6 Astra

The $20 workload: when GPT-6 Astra earns its premium

Official pricing, independent tests, safety caveats and a direct comparison with GPT-5.6 Sol.

Read the GPT-6 Astra verdict

Agenturen & Labs

The Verge AIA hacker broke into John Deere and SharkNinja with an army of AI agents n8n BlogRun your n8n AI workflows on NVIDIA DGX Spark, with no cloud API SiliconANGLE AIMotive Technologies revs up new AI Agents to automate fleet management The DecoderStripe says AI agent purchases are adding up, but it won't share the numbers The Guardian TechnologyAI firm Anthropic to commit 5% of shares to fund for ‘public good’The Verge AIDemocrats think cracking down on AI could give them an edge The DecoderAnthropic plans fund for workers threatened by AI, Musk counters with SpaceX shares for children BBC TechnologyMusk and Anthropic pledge public share giveaways MIT Technology Review AIAI researchers stranded Claude on a virtual desert island. It started to “play.”SiliconANGLE AIPax8 launches 8os to let IT partners govern AI agents and 8bot as a managed AI co-worker BBC TechnologyChina's growth accelerates as AI boom helps boost exports arXiv cs.CLEmbeddings Are Overpriced? A Frozen 16 MB Vocabulary for Small and Ternary Language Models arXiv cs.CLWhere Prefill Expert Reduction Hurts MoE Models: A Few Prompt Tokens Carry Most of the Short-Answer F1 Loss arXiv cs.LGWhen is a large language model a good calibration agent? Physics-labeled prompting for budgeted engineering simulator calibration arXiv cs.LGOn the SPOT: Enhancing Speculative Decoding for Diffusion Language Models via Commit-Order Prediction The Guardian TechnologyKylie Minogue songs already used to train AI as tech firms ‘strong-arm’ Australian government, music body warns Simon Willisonllm-tools-datasette 0.1.1 Simon Willisonllm-echo 0.5 The Verge AIA hacker broke into John Deere and SharkNinja with an army of AI agents n8n BlogRun your n8n AI workflows on NVIDIA DGX Spark, with no cloud API SiliconANGLE AIMotive Technologies revs up new AI Agents to automate fleet management The DecoderStripe says AI agent purchases are adding up, but it won't share the numbers The Guardian TechnologyAI firm Anthropic to commit 5% of shares to fund for ‘public good’The Verge AIDemocrats think cracking down on AI could give them an edge The DecoderAnthropic plans fund for workers threatened by AI, Musk counters with SpaceX shares for children BBC TechnologyMusk and Anthropic pledge public share giveaways MIT Technology Review AIAI researchers stranded Claude on a virtual desert island. It started to “play.”SiliconANGLE AIPax8 launches 8os to let IT partners govern AI agents and 8bot as a managed AI co-worker BBC TechnologyChina's growth accelerates as AI boom helps boost exports arXiv cs.CLEmbeddings Are Overpriced? A Frozen 16 MB Vocabulary for Small and Ternary Language Models arXiv cs.CLWhere Prefill Expert Reduction Hurts MoE Models: A Few Prompt Tokens Carry Most of the Short-Answer F1 Loss arXiv cs.LGWhen is a large language model a good calibration agent? Physics-labeled prompting for budgeted engineering simulator calibration arXiv cs.LGOn the SPOT: Enhancing Speculative Decoding for Diffusion Language Models via Commit-Order Prediction The Guardian TechnologyKylie Minogue songs already used to train AI as tech firms ‘strong-arm’ Australian government, music body warns Simon Willisonllm-tools-datasette 0.1.1 Simon Willisonllm-echo 0.5

Dieses Desk

DeskGrok Bot review: access, cost, risks and alternatives DeskChatGPT Work vs Claude Cowork: Best Agent for Knowledge Work DeskManus vs Genspark: Which Super Agent Produces Better Value?DeskBest Autonomous AI Agents: Manus, Genspark, ChatGPT Work and More DeskManus Pricing and Credit Calculator: What Each Completed Task Costs DeskAgentic AI: permissions as a line item DeskGrok Bot review: access, cost, risks and alternatives DeskChatGPT Work vs Claude Cowork: Best Agent for Knowledge Work DeskManus vs Genspark: Which Super Agent Produces Better Value?DeskBest Autonomous AI Agents: Manus, Genspark, ChatGPT Work and More DeskManus Pricing and Credit Calculator: What Each Completed Task Costs DeskAgentic AI: permissions as a line item

Wonach Leute den Preis suchen

Mit einer echten Kaufentscheidung beginnen

Das sind die Kostenfragen, die Käufer an dieses Desk bringen. Öffnen Sie die, die zu dem passt, was Sie zahlen wollen.

Frische Briefings

Alle öffentlichen Guides

2026-10-15 · 8 min read

Codex Cloud vs Local Cost: GPT-5.3-Codex Pricing Compared

Prices the GPT-5.3-Codex API against ChatGPT subscription and self-hosted usage patterns, using the official 12-step reference run to show what a coding agent actually costs.

2026-10-15 · 11 min read

Agent Evals Cost: Per-Run and Monthly Pricing Breakdown

Prices a 12-step agent eval trace across 12 models, cached and uncached, with monthly cost scenarios for 1k, 10k, and 100k eval runs.

2026-10-12 · 11 min read

Claude Code API vs Max Plan Cost: The Break-Even Math

Whether the Claude Code Max subscription beats paying Anthropic API rates for agent runs, worked out from Sonnet 5, Opus 5 and Haiku 4.5 per-run costs.

2026-10-01 · 12 min read

Cheapest LLM for AI Agents: 30 Models Ranked by Cost Per Run

Every current API model ranked by the cost of the same 12-step tool-calling run, a reliability tier list, the route-cheap-escalate-on-failure pattern with code, and when the flagship is cheaper per successful task.

2026-10-01 · 12 min read

AI Model Routing: How to Cut Costs Without Losing Quality

How to route AI requests to cheaper models intelligently — the technique used by companies spending millions on AI to stay profitable.

2026-10-01 · 8 min read

ChatGPT vs Claude: Which Costs Less in 2026?

A precise cost comparison between ChatGPT and Claude — $20 subscriptions on both sides, and per-token API rates — to find the cheaper option for your actual usage.

Öffentliche Vergleiche

Alle öffentlichen Vergleiche

2026-10-17

ChatGPT vs Cursor Cost in 2026: $20 Plus vs Cursor's $8 to $200 Tiers

ChatGPT Plus is still $20/month. Cursor no longer has a $20 Pro plan for new users: it sells Plus at $8, Premium at $30, Super at $100 and Ultra at $200, with Claude and GPT models billed on top. Here is the real monthly cost of each tier and which one to pay for.

2026-10-17

AI Subscription vs API Cost in 2026: Find Your Break-Even

A numeric cost comparison of a $20/month ChatGPT Plus or Claude Pro subscription against GPT-6 Sol, GPT-5.5, Claude Sonnet 5 and the cheap API tiers, with a worked break-even table you can check against your own token mix.

2026-10-17

Claude vs Cursor for Coding in 2026: API, Claude Pro or a Cursor Seat?

Claude Sonnet 5 via API ($2/$10 per 1M tokens), Claude Code on Claude Pro ($20/month) and a Cursor Premium seat ($30/month plus metered model spend) compared for professional software development, with real monthly numbers.

2026-09-11

Microsoft Copilot vs ChatGPT Plus: Which Is Better for Office Work?

Microsoft Copilot Pro is $30/month; ChatGPT Plus is $20. Both run OpenAI's GPT-5 models — the question is whether native Word, Excel and Outlook integration is worth $10 more.

2026-09-11

DeepSeek vs ChatGPT vs Claude: The $0 vs $20/Month vs API Decision

DeepSeek's chat is free. ChatGPT Plus and Claude Pro both cost $20/month. On the API, DeepSeek V4 Flash at $0.05/$0.16 per 1M tokens undercuts GPT-5.5 and Claude Sonnet 5 by 49x or more. A decision matrix for which surface fits your usage.

2026-09-11

Claude Free vs Pro: Is Claude Pro Worth $20/Month?

Claude's free tier gives you Claude Sonnet 5 with limits. Claude Pro costs $20/month and adds higher limits, Opus 5 and Claude Code. Here is what the upgrade actually adds and who should pay.

Mit der Aufgabe beginnen

Coding

Editors, agents, accepted diffs

Writing

Drafts that survive review

Research

Cited briefs, not busy agents

Support

Resolved tickets, not chat volume

Agents

Permissions as a line item

Design

Approved assets, not generations

Training desk

Learn the model, then price it

140 free lessons across 25 courses. Every one ends in what the model costs and the cheapest way to get the same result — no sign-up, progress saved in your browser.

All tutorials

New to AI? Start here

AI From Zero6 lessons · 30 min · no sign-upSix short lessons and no account needed to try the examples. Finish one real task with AI today, learn when not to trust it, and find out whether you need to pay for anything at all.Which AI Should I Use?6 lessons · 36 min · no sign-upSix needs, not six brands. For each one: two or three candidate tools, a prompt you can run to feel the difference, and a plain verdict on where to start and when to move.How to Talk to AI5 lessons · 28 min · no sign-upOne three-part habit, five short lessons, no vocabulary to memorise. Get a usable answer on the first try more often, and know exactly what to say when you do not.10 Things With AI5 lessons · 32 min · no sign-upTen small, useful jobs — an email, a summary, a plan, a spreadsheet question — each one done in two minutes with a prompt you can paste, and a note on which ones a free account handles.When AI Is Wrong6 lessons · 35 min · no sign-upA fluent answer is not a checked answer. Six short lessons that teach you why a chat model can be confidently wrong, how to spot it, and the one thirty-second check that works every time.AI With PDFs5 lessons · 33 min · no sign-upUpload a document, ask it questions, summarise it to the length you want, pull out the action items, compare two versions, and know exactly where a free plan stops being enough.

One product, in depth

First Output 1016 lessons · 72 min · BeginnerSix lessons, six things you actually make. Start with one API call costing a fraction of a cent, finish with a working tool and a routing rule that keeps your bill boring.Claude 1018 lessons · 81 min · BeginnerEverything Anthropic's own Claude 101 teaches — plus what each plan and model actually costs, and the cheapest way to get the same result.ChatGPT 1018 lessons · 81 min · BeginnerLearn ChatGPT from zero, and learn exactly what each plan and model costs before you pay for one.Gemini 1018 lessons · 80 min · Beginner → IntermediateLearn what Gemini is, which model to pick, and what each plan and API call really costs before you pay Google a cent.DeepSeek 1018 lessons · 79 min · Beginner → IntermediateLearn the free chat app, the open-weight model family and the OpenAI-compatible API — and what each one actually costs.Mistral 1018 lessons · 72 min · BeginnerLe Chat, the Mistral model families, open weights, and what each one actually costs.Llama 1018 lessons · 79 min · Beginner → IntermediateOpen-weight models you can rent by the token or run yourself. Learn what that actually costs.Grok 1018 lessons · 73 min · Beginner → IntermediateLearn what Grok is, which model to pick, what it really costs, and how to keep the bill small.

Still worth opening

Updated recently

Pricing changelog

Compare · Oct 17, 2026ChatGPT vs Cursor Cost in 2026: $20 Plus vs Cursor's $8 to $200 TiersThe old 'same $20' framing is dead. ChatGPT Plus still costs $20/month flat. Cursor's new plans are Plus at $8 (Grok models only), Premium at $30, Super at $100 and Ultra at $200, and on Premium and above every Claude, GPT or Gemini request is billed separately at API list price. A developer on Premium who sends a moderate 3M input and 600K output tokens a month to Claude Sonnet 5.5 pays about $42, not $30. For a generalist, ChatGPT Plus at $20 is still the cheaper and broader buy. For a full-time developer, Cursor Premium at a realistic $40 to $50 a month is now the honest comparison.Compare · Oct 17, 2026AI Subscription vs API Cost in 2026: Find Your Break-EvenThis is the numeric cost page, and the API is cheaper for most people. At GPT-6 Sol's $2/1M input and $10/1M output, a $20 ChatGPT Plus subscription breaks even at about 5M tokens a month on a 3:1 input-to-output mix. Claude Sonnet 5 at $2/$10 lands in the same place. Route routine work to GPT-5.4 mini ($0.75/$4.50) and the break-even moves past 11M tokens; on GPT-6 Luna ($0.10/$0.50) it is about 100M. Subscriptions win on convenience for daily chat; APIs win on cost for automation, agents and high-volume work.Best of · Oct 17, 2026Cheapest AI Code Editors and Assistants on a BudgetAI coding tools ranked by what you actually pay: free editor tiers, the $20 subscriptions, and the cheapest APIs that still handle agent tool calls.Best of · Oct 17, 2026Best AI Subscriptions Under $20/MonthEvery mainstream AI subscription at $20 or less, ranked by what it costs each month: Google AI Plus at $4.99, ChatGPT Go and Cursor Plus at $8, then the $20 plans from ChatGPT Plus, Claude Pro, Perplexity Pro and Google AI Pro. Cursor Pro no longer makes the cut.Guide · Oct 15, 2026Codex Cloud vs Local Cost: GPT-5.3-Codex Pricing ComparedPrices the GPT-5.3-Codex API against ChatGPT subscription and self-hosted usage patterns, using the official 12-step reference run to show what a coding agent actually costs.Guide · Oct 15, 2026Agent Evals Cost: Per-Run and Monthly Pricing BreakdownPrices a 12-step agent eval trace across 12 models, cached and uncached, with monthly cost scenarios for 1k, 10k, and 100k eval runs.Pricing · Oct 1, 2026OpenRouter Pricing: Fees, Business Plan, Batch and Routers (2026)OpenRouter bills provider rates plus a 5.5% credit fee, or 8% on the new Business plan. See Batch API discounts, BYOK, routers, shell and in-region costs.Pricing · Oct 1, 2026Jev Pricing: TypeSafe's Decision Model Cost per Million DecisionsJev costs $0.042 per million input tokens and output is free. See cost per million decisions, OpenRouter fees, and when Jev beats a cheap LLM.

Wie dieses Desk ehrlich bleibt

01Katalog-API-Preise werden über den Tag aktualisiert. Ist eine Zahl älter als zwei Wochen, markieren wir sie als veraltet statt aktuell wirken zu lassen.
02News-Briefings verlassen das öffentliche Desk nach drei Wochen. Dauerhafte Kaufseiten bleiben gelistet, solange der API-Katalog Live oder Recent ist. Wir setzen keine Daten vor, nur um aktuell zu wirken.
03Die Fragen oben sind die Käufe, die Menschen wirklich abwägen. Neue Vergleiche erscheinen, solange sie aktuell sind.

Partner picks

Claude Replit Make Taskade Fireflies.ai Pressmaster.ai Murf AI

Affiliate links — we may earn a commission

Den echten Preis der KI kennen, bevor Sie zahlen.

Mit einer echten Kaufentscheidung beginnen

Frische Briefings

Öffentliche Vergleiche

Learn the model, then price it

Updated recently

Wie dieses Desk ehrlich bleibt

Stop overpaying for AI tools