AI Coding·Plans
EN
2026 Vergleich · 15 Anbieter · mehr als codingplan.org

AI-Coding-Abos & Agenten im Vergleich

Alle großen AI-Coding-Subscriptions auf einer Seite: Preise, Modelle, Kontext-Limits, Free-Tiers und Empfehlungen. Mehr Anbieter als jedes andere Vergleichsportal.

Zum Vergleich ▾

Schnellvergleich

Klick auf eine Spalte zum Sortieren. Preise sind Richtwerte in USD/Monat und können sich häufig ändern.

36 von 36 Anbietern
Anbieter ↕ Ab /Monat ↕ Top-Modelle Kontext Free-Tier Bestens für Link
Anthropic
 $20 Claude Opus 4Claude Sonnet 4 200K tokens ✓ Power-users wanting the strongest reasoning model in-terminal Website →
Anysphere
 $20 Claude Sonnet 4GPT-5Gemini 2.5 Pro model-dependent — Devs who want AI inside a familiar editor Website →
GitHub/Microsoft
 $10 GPT-5Claude Sonnet 4Gemini 2.5 Pro model-dependent ✓ Teams already in the GitHub ecosystem Website →
Google
 Kostenlos Gemini 2.5 ProGemini 2.5 Flash 1M–2M tokens ✓ Long-context files and Google Cloud/Firebase users Website →
Codeium/Windsurf
 $15 Claude Sonnet 4GPT-5SWE-1 model-dependent ✓ Teams wanting an agentic IDE alternative to Cursor Website →
Replit
 $25 Claude Sonnet 4GPT-5Replit-internal model-dependent — Zero-setup app building + hosting in one place Website →
open source (community)
 Kostenlos Claude Sonnet 4GPT-5Gemini 2.5 ProQwen3 model-dependent ✓ Devops who want a transparent, hackable agent Website →
open source (community)
 Kostenlos Claude Sonnet 4GPT-5Gemini 2.5 ProQwen3 model-dependent ✓ Power-users wanting custom modes & subagents Website →
Tabnine (Braintrust)
 $9 Tabnine proprietaryon-prem models n/a ✓ Enterprises with strict data/IP requirements Website →
Sourcegraph
 $9 Claude Sonnet 4GPT-5Gemini 2.5 Pro large repo context ✓ Large, multi-repo codebases needing code search + AI Website →
z.ai (Zhipu)
 $3 GLM-4.6GLM-4.5-Air 200K tokens ✓ Cost-effective flat-rate agentic coding Website →
MiniMax
 Kostenlos MiniMax M2abab 1M+ tokens ✓ High-context, cost-effective coding via API Website →
Moonshot AI
 $0 Kimi K2Moonshot-v1 2M tokens ✓ Feeding entire repos/docs into context Website →
Alibaba Cloud
 Kostenlos Qwen3-CoderQwen2.5-Coder 256K–1M tokens ✓ Self-hosters and open-weight fans Website →
DeepSeek
 $0 DeepSeek-V3DeepSeek-R1DeepSeek-Coder-V2 128K tokens ✓ Budget-conscious coding with frontier-ish quality Website →
Amazon Web Services
 Kostenlos Amazon QAnthropic Claude model-dependent ✓ AWS shops wanting inline chat, agents & security scans Website →
Augment
 $20 Augment proprietaryfrontier models large codebase context — Large, complex codebases needing deep repo context Website →
Continue (open source)
 Kostenlos Claude Sonnet 4GPT-5Gemini 2.5 ProQwen3 model-dependent ✓ Devs wanting a hackable, provider-agnostic copilot Website →
Paul Gauthier (open source)
 Kostenlos Claude Sonnet 4GPT-5DeepSeek model-dependent ✓ Terminal-first devs who want git-native AI edits Website →
Refact
 Kostenlos Refact-1.6frontier models model-dependent ✓ Teams wanting self-hostable open-weight coding LLMs Website →
Bito
 Kostenlos GPT-5Claude Sonnet 4Gemini 2.5 Pro model-dependent ✓ Devs wanting chat + completions + code review in one Website →
Z.ai (Zhipu) / Tsinghua
 Kostenlos CodeGeeX-4open weights model-dependent ✓ Devs wanting a free, open-source multilingual copilot Website →
Blackbox
 Kostenlos Blackbox proprietaryfrontier models model-dependent ✓ Quick code answers, snippets & autocomplete on a budget Website →
Sourcery / Sourcegraph
 Kostenlos frontier models repo context ✓ Teams wanting automated refactoring & code review Website →
Phind
 Kostenlos Phindfrontier models model-dependent ✓ Developers researching answers & code online Website →
StackBlitz
 Kostenlos Claude Sonnet 4GPT-5Gemini 2.5 Pro model-dependent ✓ Building & shipping fullstack apps from a prompt Website →
Vercel
 Kostenlos GPT-5Claude Sonnet 4Gemini 2.5 Pro model-dependent ✓ Frontend devs generating UI & Next.js components fast Website →
Lovable
 Kostenlos Claude Sonnet 4GPT-5 model-dependent ✓ Building & shipping fullstack apps with minimal coding Website →
Zed Industries
 Kostenlos Claude Sonnet 4GPT-5Gemini 2.5 Pro model-dependent ✓ Devs wanting a fast native editor with AI built in Website →
Cognition
 $20 Devin (Cognition) agentic sandbox — Offloading well-scoped engineering tasks to an autonomous agent Website →
Warp
 Kostenlos GPT-5Claude Sonnet 4Gemini 2.5 Pro model-dependent ✓ Terminal-first devs wanting AI command help & agents Website →
Qodo
 Kostenlos frontier models repo context ✓ Teams focused on tests, coverage & AI code review Website →
BigCode / Hugging Face
 Kostenlos StarCoder2StarCoder2-Instruct 16K tokens ✓ Self-hosters wanting a free open-weight code model Website →
All Hands AI (open source)
 Kostenlos Claude Sonnet 4GPT-5open models model-dependent ✓ Devs wanting a transparent autonomous agent they control Website →
Pythagora (open source)
 Kostenlos GPT-5Claude Sonnet 4 model-dependent ✓ Devops experimenting with prompt-to-app generation locally Website →
Void (open source)
 Kostenlos Claude Sonnet 4GPT-5local models model-dependent ✓ Privacy-focused devs wanting an open-source AI IDE Website →

Anbieter vergleichen & teilen

2–3 Anbieter auswählen — Vergleichsansicht + teilbarer Link (URL-Parameter).

Anbieter zum Vergleichen auswählen (max. 3)
NEU · TCO-Rechner

Total Cost of Ownership: Welches Abo passt zu dir?

Dein Nutzungsprofil eingeben — wir schätzen die monatlichen Gesamtkosten pro Anbieter aus providers.json und empfehlen die günstigste passende Wahl. Logik & Faktoren vollständig einsehbar.

Wie viele AI-Anfragen (Codierung/Chat) sendet ihr täglich?
Empfehlung bevorzugt Open-Source-Clients (Cline, Roo, Qwen, DeepSeek).

⚠ Schätzung anhand dokumentierter Heuristiken (Preis-Tiers, Billing, Usage-Annahmen). Keine Garantie — immer auf der Anbieter-Seite prüfen.

Detaillierte Pläne

Pro Anbieter: Modelle, Limits, Stärken & Schwächen.

Claude Code

Anthropic
Agentic terminal coding on Claude Sonnet/Opus.
Ab /Monat$20
Kontext200K tokens
Billingsubscription + usage
Claude Opus 4Claude Sonnet 4
Limits
Pro: ~45 msgs/5h; Max: 5x–20x Pro usage
Free-Tier: limited free credits on console
  • Top-tier reasoning
  • Strong agentic tool use
  • First-class CLI
  • Usage caps can hit fast on Pro
  • Premium pricing
Zu Claude Code → Details →

Cursor

Anysphere
AI-first VS Code fork with Composer + Tab.
Ab /Monat$20
Kontextmodel-dependent
Billingsubscription
Claude Sonnet 4GPT-5Gemini 2.5 Pro
Limits
Pro: 500 fast premium requests/mo
Free-Tier: Hobby: 50 fast + slow requests/mo
  • Best-in-class autocomplete (Tab)
  • Multi-file Composer
  • Model choice
  • Credit-based fast requests
  • Editor lock-in
Zu Cursor → Details →

GitHub Copilot

GitHub/Microsoft
The default AI pair programmer for the IDE.
Ab /Monat$10
Kontextmodel-dependent
Billingsubscription
GPT-5Claude Sonnet 4Gemini 2.5 Pro
Limits
Generous multi-model usage; agent mode
Free-Tier: Free tier for students/maintainers + limited monthly
  • Deep GitHub integration
  • Multi-model
  • Enterprise features
  • Less aggressive agentic loop than Cursor
  • Usage controls vary by plan
Zu GitHub Copilot → Details →

Gemini Code Assist

Google
Gemini-powered completions + agent in IDE & Firebase.
Ab /MonatKostenlos
Kontext1M–2M tokens
Billingfreemium + subscription
Gemini 2.5 ProGemini 2.5 Flash
Limits
Free: 180 chat/agent + 6,000 completions/day
Free-Tier: yes — generous free individual tier
  • Huge context window
  • Strong free tier
  • GCP integration
  • Best within Google ecosystem
  • Agent maturity still catching up
Zu Gemini Code Assist → Details →

Windsurf

Codeium/Windsurf
Agentic IDE with Cascade flows.
Ab /Monat$15
Kontextmodel-dependent
Billingsubscription + credits
Claude Sonnet 4GPT-5SWE-1
Limits
Pro: flexible flow-based credits
Free-Tier: yes — limited free tier
  • Cascade multi-step agent
  • Strong enterprise story
  • Own SWE-1 model
  • Newer ecosystem
  • Credit accounting can be opaque
Zu Windsurf → Details →

Replit Agent

Replit
Build & deploy whole apps from a prompt in the cloud.
Ab /Monat$25
Kontextmodel-dependent
Billingsubscription + usage
Claude Sonnet 4GPT-5Replit-internal
Limits
Monthly checkpoints + agent usage
Free-Tier: limited Starter tier
  • Full cloud dev environment
  • Deploy-from-prompt
  • Great for prototyping
  • Less control than local IDE
  • Usage can add up
Zu Replit Agent → Details →

Cline

open source (community)
Open-source autonomous coding agent for VS Code.
Ab /MonatKostenlos
Kontextmodel-dependent
BillingBYO-key (usage)
Claude Sonnet 4GPT-5Gemini 2.5 ProQwen3
Limits
whatever your API key allows
Free-Tier: free software (model usage billed separately)
  • Fully open source
  • Works with any API provider
  • Transparent plans
  • You manage & pay for model usage
  • Less polish than commercial IDEs
Zu Cline → Details →

Roo Code

open source (community)
Open-source VS Code agent fork with custom modes.
Ab /MonatKostenlos
Kontextmodel-dependent
BillingBYO-key (usage)
Claude Sonnet 4GPT-5Gemini 2.5 ProQwen3
Limits
whatever your API key allows
Free-Tier: free software (model usage billed separately)
  • Custom modes/subagents
  • Open source
  • Highly configurable
  • You supply & pay for the model
  • Steeper config
Zu Roo Code → Details →

Tabnine

Tabnine (Braintrust)
Privacy-first completions; self-hosted models.
Ab /Monat$9
Kontextn/a
Billingsubscription
Tabnine proprietaryon-prem models
Limits
per-seat
Free-Tier: basic free tier
  • Strong privacy/SOC2 story
  • Self-hosted option
  • Enterprise focus
  • Weaker than frontier models on agentic tasks
  • Smaller community
Zu Tabnine → Details →

Cody

Sourcegraph
AI coding assistant that knows your whole codebase.
Ab /Monat$9
Kontextlarge repo context
Billingsubscription
Claude Sonnet 4GPT-5Gemini 2.5 Pro
Limits
multi-model with repo-graph context
Free-Tier: limited free tier
  • Whole-codebase context via code graph
  • Multi-model
  • Enterprise search
  • Best value is in big repos
  • Editor experience varies
Zu Cody → Details →

GLM Coding Plan (z.ai)

z.ai (Zhipu)
GLM-4.6 agentic coding with a flat-rate coding plan.
Ab /Monat$3
Kontext200K tokens
Billingflat subscription (prompt quota)
GLM-4.6GLM-4.5-Air
Limits
prompt-quota based, not token-priced
Free-Tier: promotional free credits
  • Flat-rate pricing
  • Strong agentic GLM models
  • Low cost
  • Newer ecosystem in West
  • Quota can cap heavy bursts
Zu GLM Coding Plan (z.ai) → Details →

MiniMax M-Series

MiniMax
Frontier Chinese model with strong coding/agentic results.
Ab /MonatKostenlos
Kontext1M+ tokens
Billingusage
MiniMax M2abab
Limits
API rate limits
Free-Tier: promotional free tier
  • Huge context
  • Competitive benchmarks
  • Low usage pricing
  • Thin native tooling
  • Mostly API-only
Zu MiniMax M-Series → Details →

Kimi (Moonshot)

Moonshot AI
Ultra-long-context model good for big-codebase tasks.
Ab /Monat$0
Kontext2M tokens
Billingusage
Kimi K2Moonshot-v1
Limits
API rate limits
Free-Tier: free web/app tier for chat
  • Massive context
  • Agentic K2
  • Cheap long-context
  • API-only coding UX
  • Less Western tooling
Zu Kimi (Moonshot) → Details →

Qwen Coder (Alibaba)

Alibaba Cloud
Top open-weight coding models, API or self-host.
Ab /MonatKostenlos
Kontext256K–1M tokens
Billingusage / self-host
Qwen3-CoderQwen2.5-Coder
Limits
hardware or API limits
Free-Tier: open-weight models are free
  • Open weights, self-hostable
  • Strong coding benchmarks
  • Cheap at scale
  • DIY tooling/integration
  • Self-host ops overhead
Zu Qwen Coder (Alibaba) → Details →

DeepSeek Coder

DeepSeek
Strong reasoning coding model at very low API cost.
Ab /Monat$0
Kontext128K tokens
Billingusage
DeepSeek-V3DeepSeek-R1DeepSeek-Coder-V2
Limits
API rate limits
Free-Tier: promotional free credits
  • Extremely cheap
  • Good reasoning
  • Open weights available
  • Capacity/cap throttling at peak
  • Less polished tooling
Zu DeepSeek Coder → Details →

Amazon Q Developer

Amazon Web Services
AWS-native AI coding assistant (formerly CodeWhisperer).
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + subscription
Amazon QAnthropic Claude
Limits
Free: monthly usage caps; Pro: higher limits
Free-Tier: yes — perpetual free tier
  • Deep AWS integration
  • Security scanning
  • Generous free tier
  • AWS-centric
  • New signups sunset after May 15, 2026
Zu Amazon Q Developer → Details →

Augment Code

Augment
Context-engine IDE agent built for large enterprise codebases.
Ab /Monat$20
Kontextlarge codebase context
Billingsubscription + credits
Augment proprietaryfrontier models
Limits
credit-based usage since Oct 2025
Free-Tier: limited trial
  • Huge codebase context
  • Strong enterprise story
  • Fast completions
  • Credit-based costs can climb
  • Newer vendor
Zu Augment Code → Details →

Continue.dev

Continue (open source)
Open-source AI coding autopilot for VS Code & JetBrains.
Ab /MonatKostenlos
Kontextmodel-dependent
BillingBYO-key (usage)
Claude Sonnet 4GPT-5Gemini 2.5 ProQwen3
Limits
whatever your API key allows
Free-Tier: free software (model usage billed separately)
  • Fully open source
  • Any local or cloud model
  • VS Code + JetBrains
  • You manage & pay for model usage
  • Less polish than paid IDEs
Zu Continue.dev → Details →

Aider

Paul Gauthier (open source)
AI pair programming in the terminal that edits your git repo.
Ab /MonatKostenlos
Kontextmodel-dependent
BillingBYO-key (usage)
Claude Sonnet 4GPT-5DeepSeek
Limits
whatever your API key allows
Free-Tier: free software (model usage billed separately)
  • Git-native workflow
  • Works with any API provider
  • Lightweight CLI
  • CLI only
  • You supply & pay for the model
Zu Aider → Details →

Refact.ai

Refact
Open-weight LLMs + hosted agent with autonomous mode.
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + credits
Refact-1.6frontier models
Limits
Free: daily agent cap; Pro: coin-based
Free-Tier: yes — personal free tier
  • Open-weight, self-hostable models
  • Autonomous agent
  • Strong free tier
  • Smaller ecosystem
  • Coin accounting takes getting used to
Zu Refact.ai → Details →

Bito

Bito
AI assistant with chat, completions, CLI & code reviews.
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + subscription
GPT-5Claude Sonnet 4Gemini 2.5 Pro
Limits
Free: 250 completions/mo
Free-Tier: yes — free single seat
  • All-in-one assistant
  • AI code reviews
  • CLI companion
  • Usage-based add-ons
  • Lesser-known brand
Zu Bito → Details →

CodeGeeX

Z.ai (Zhipu) / Tsinghua
Multilingual open-source coding assistant (100+ languages).
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + subscription
CodeGeeX-4open weights
Limits
Free: limited credits; Pro: 500 credits
Free-Tier: yes — generous free tier
  • Open weights, self-hostable
  • 100+ languages
  • Strong free tier
  • Behind frontier models on hard tasks
  • Smaller community
Zu CodeGeeX → Details →

Blackbox AI

Blackbox
AI coding assistant & code-aware search for devs.
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + subscription
Blackbox proprietaryfrontier models
Limits
Free: unlimited chat, basic completions
Free-Tier: yes — unlimited chat
  • Very cheap Pro tier
  • Unlimited free chat
  • Code search
  • Less agentic depth
  • Quality below frontier models
Zu Blackbox AI → Details →

Sourcery

Sourcery / Sourcegraph
AI refactoring & code-review agent (strong for Python).
Ab /MonatKostenlos
Kontextrepo context
Billingfreemium + subscription
frontier models
Limits
Free: 20 issue investigations/mo
Free-Tier: free for open-source repos
  • Automated refactoring
  • AI code review
  • Strong Python support
  • Narrow refactoring focus
  • Now folded into Sourcegraph
Zu Sourcery → Details →

Phind

Phind
Answer engine for developers with code search & generation.
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + subscription
Phindfrontier models
Limits
Free: daily search cap
Free-Tier: yes — free tier
  • Great dev search
  • Cited answers
  • Strong free tier
  • Search-focused, not a full IDE agent
  • Credits can run out
Zu Phind → Details →

Bolt.new

StackBlitz
Prompt-to-app builder running fullstack in the browser.
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + tokens
Claude Sonnet 4GPT-5Gemini 2.5 Pro
Limits
token-based; Pro: 10M tokens/mo
Free-Tier: yes — daily token allowance
  • Deploy-from-prompt
  • Zero-setup browser IDE
  • Fast prototyping
  • Token costs on big apps
  • Less control than local dev
Zu Bolt.new → Details →

v0 (Vercel)

Vercel
Generative UI — build React/Next.js components from prompts.
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + credits
GPT-5Claude Sonnet 4Gemini 2.5 Pro
Limits
credit-based
Free-Tier: yes — limited free credits
  • First-class Next.js/React output
  • Vercel deploy integration
  • Fast UI gen
  • Frontend/UI focused
  • Credits run out on heavy use
Zu v0 (Vercel) → Details →

Lovable

Lovable
Conversational fullstack app builder for non-coders & devs.
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + credits
Claude Sonnet 4GPT-5
Limits
credit-based
Free-Tier: yes — limited free credits
  • Deploy-from-prompt
  • Beginner-friendly
  • Fullstack output
  • Credit-based
  • Less granular control
Zu Lovable → Details →

Zed (AI)

Zed Industries
High-performance editor with built-in AI & edit predictions.
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + usage
Claude Sonnet 4GPT-5Gemini 2.5 Pro
Limits
Pro: $5 tokens included, then usage
Free-Tier: free trial
  • Blazing-fast native editor
  • Inline AI edits
  • Token-based flexibility
  • Smaller ecosystem
  • Token usage adds up
Zu Zed (AI) → Details →

Devin

Cognition
Autonomous AI software engineer handling whole tasks end-to-end.
Ab /Monat$20
Kontextagentic sandbox
Billingsubscription + usage (ACU)
Devin (Cognition)
Limits
ACU-based; ~$9/h of Devin time
Free-Tier: no — paid only
  • Truly autonomous end-to-end agent
  • Parallel tasks
  • PR-ready output
  • Expensive at scale
  • Needs well-scoped tasks
Zu Devin → Details →

Warp

Warp
Agentic terminal with built-in AI for commands & workflows.
Ab /MonatKostenlos
Kontextmodel-dependent
Billingfreemium + subscription
GPT-5Claude Sonnet 4Gemini 2.5 Pro
Limits
Free: no AI credits; Pro: ~1,500 credits/mo
Free-Tier: yes — core terminal free
  • AI-native terminal
  • Frontier model access
  • Workflow automation
  • Recent price hikes
  • Terminal-only focus
Zu Warp → Details →

Qodo (ex-CodiumAI)

Qodo
AI code review, test generation & quality focused agents.
Ab /MonatKostenlos
Kontextrepo context
Billingfreemium + credits
frontier models
Limits
Free: ~75 PRs/mo; paid credit-based
Free-Tier: yes — generous free tier
  • Strong test generation
  • AI code review
  • Coverage focus
  • Quality/coverage niche
  • Credit-based
Zu Qodo (ex-CodiumAI) → Details →

StarCoder2 (BigCode)

BigCode / Hugging Face
Open-weight code LLM family for self-hosted completion.
Ab /MonatKostenlos
Kontext16K tokens
Billingself-host / usage
StarCoder2StarCoder2-Instruct
Limits
hardware or API limits
Free-Tier: open-weight models are free
  • Open weights, self-hostable
  • Commercially usable
  • Free
  • DIY integration
  • Lags frontier models
Zu StarCoder2 (BigCode) → Details →

OpenHands (ex-OpenDevin)

All Hands AI (open source)
Open-source autonomous agent that codes, browses & runs commands.
Ab /MonatKostenlos
Kontextmodel-dependent
BillingBYO-key (usage)
Claude Sonnet 4GPT-5open models
Limits
whatever your API key allows
Free-Tier: free software (model usage billed separately)
  • Fully open source
  • Autonomous multi-tool agent
  • Self-hostable
  • You supply & pay for the model
  • Setup required
Zu OpenHands (ex-OpenDevin) → Details →

GPT-Pilot

Pythagora (open source)
Open-source research agent that builds whole apps step by step.
Ab /MonatKostenlos
Kontextmodel-dependent
BillingBYO-key (usage)
GPT-5Claude Sonnet 4
Limits
whatever your API key allows
Free-Tier: free software (model usage billed separately)
  • Fully open source
  • Prompt-to-app
  • Runs locally
  • Research-grade polish
  • You supply & pay for the model
Zu GPT-Pilot → Details →

Void

Void (open source)
Open-source, privacy-first Cursor alternative IDE.
Ab /MonatKostenlos
Kontextmodel-dependent
BillingBYO-key (usage)
Claude Sonnet 4GPT-5local models
Limits
whatever your API key allows
Free-Tier: free software (model usage billed separately)
  • Fully open source
  • Local-model support
  • Privacy-first
  • Young project
  • You supply & pay for the model
Zu Void → Details →

Häufige Fragen (FAQ)

Welches AI-Coding-Abo ist das beste?

Das hängt vom Anwendungsfall ab: Für maximale Reasoning-Qualität im Terminal ist Claude Code führend; für ein fertiges Editor-Erlebnis Cursor oder Windsurf; für Budget-Bewusste GLM/z.ai, DeepSeek oder Qwen; für Self-Hoster die Open-Source-Agenten Cline & Roo Code.

Gibt es wirklich kostenlose Optionen?

Ja. Gemini Code Assist, Cline, Roo Code, Qwen, DeepSeek, MiniMax und Kimi bieten deutliche Free-Tiers oder sind Open Source (du zahlst nur die Modell-Usage deines API-Keys).

Subscription vs. Usage-API — was lohnt sich?

Subscriptions (Cursor, Copilot, Claude Pro/Max) sind planbar und haben weiche Caps. Usage-APIs (DeepSeek, Qwen, GLM, OpenAI) sind oft billiger bei hohem Volumen, aber unvorhersehbar im Preis. Agenten wie Cline/Roo kombinieren Open-Source-Client mit deiner eigenen Usage-API.

Wie aktuell sind die Preise?

Anbieter ändern Preise und Limits häufig. Wir pflegen die Daten zentral in providers.json und aktualisieren regelmäßig — immer verifizieren auf der Anbieter-Seite vor dem Abschluss.

Was ist mit Datenschutz & Enterprise?

Tabnine und Cody (Sourcegraph) haben starke Privacy-/Self-Host-Stories; GitHub Copilot und Windsurf bieten Enterprise-Pläne mit SOC2/Admin-Controls.

Affiliate-Hinweis. Diese Seite kann Referral-/Affiliate-Links enthalten. Wir erhalten ggf. eine Vergütung, wenn du über diese Links abschließt — für dich entstehen keine Mehrkosten. Preise und Limits sind unverbindlich und ohne Gewähr (Stand 2026).