| PricePrice | Kostenlos: $0 / month (ECB conversion) | Free: $0 / month | Free: $0 / month (ECB conversion) | Pro: $233.98 / year (ECB conversion) | Premium from: $4.99 / month | SuperGrok: $30 / month | GPT-5.4 mini cached input / 1M tokens: $0.08 / per use | Haiku 4.5 cache read / 1M tokens: $0.10 / per use | Gemini Flash input / 1M tokens: $0.25 / per use | Gemini Flash Image input / 1M tokens: $0.50 / per use | Mistral Small input / 1M tokens: $0.15 / per use | Pro from: $8 / month | Hobby: $0 / month | Starter: $19 / month | Mac: $128.38 / year | Free: $0 / one-time (ECB conversion) | Basic: $29 / one-time | Free: $0 / month | Basic: $96 / year | Free: $0 / month | Free: $0 / month |
| Limits and resetsUsage limit | Exact core and tool limits can be dynamic and are shown in-product. | Pro provides at least five times free usage per session; a weekly limit also applies. | Google publishes relative usage multipliers by plan. | Consumer limits are published as weekly and monthly classes rather than fixed counts. | Messages consume different compute points depending on the bot. | X publishes increased Grok limits for consumers but no stable exact count. | Rate limits differ by model and usage tier. | Requests, tokens, and spend are limited by usage tier. | Limits apply per project and are measured as RPM, TPM, and RPD. | Free models are limited to 20 requests per minute and 50 or 1,000 per day. | API limits apply per organization and model. | Pro models generally allow 50 requests per minute and 300 per hour. | Cursor publishes approximate model-dependent usage for median use. | Free models are limited to 200 requests per hour per IP. | Setapp publishes monthly AI credits by plan. | The one-time Pro license promises unlimited transcriptions. | The free plan is limited to five conversations and ten inline uses per day. | Individual plans include tiered monthly AI credits. | Plans include 3.3 to 60 hours of Fast GPU time per month. | Paid plans include 2,000 to 50,000 generative credits monthly. | Plans include 125 one-time credits or 625 to 9,500 monthly credits. |
| Limits and resetsReset | The exact reset time is shown in-product for the account. | The session allowance resets after five hours; the weekly limit resets at a fixed account time. | Model limits use five-hour and weekly windows. | Max includes 10,000 Computer credits per month. | Plan points are granted daily and/or monthly and usually do not roll over. | No single public reset time is specified. | Requests and tokens are limited in quantized time windows. | Exact account limits depend on tier and model. | Daily request quotas reset at midnight Pacific time. | API keys expose remaining credits and reset type. | Exact values are account-specific in the admin console. | Advanced AI uses three-hour, daily, and sometimes weekly windows. | Included usage is measured monthly. | Organization daily user limits reset at midnight UTC. | Unused AI credits expire at the next billing period. | Local transcription has no subscription allowance reset. | The free inline allowance is counted daily. | Included AI credits reset monthly. | Fast GPU time refreshes with the monthly billing cycle. | Generative credits reset on the monthly billing date and do not roll over. | Standard and Pro credits expire on the billing date; Max rolls over up to one month. |
| FitStrengths | Very broad toolset for text, files, images, and research. · Codex and data analysis also make the plan useful for developers. · Apps for web, macOS, Windows, iOS, and Android. | Excellent writing quality and work with large documents. · Claude Code and integrations support demanding development workflows. · Usage windows and resets are documented comparatively well. | Deep integration with Google services and Workspace. · Large context windows and strong multimodal features. · Relative plan and usage allowances are officially documented. | Citations are central to the product experience. · Multiple capable models are available through one subscription. · Strong fit for web research and evidence-oriented writing. | Many text, image, and media models in one interface. · A points system makes differently priced models usable in one plan. · Web and mobile access with large-context options. | Real-time access to content from the X ecosystem. · Strong reasoning and coding orientation. · Image and research features in one chat product. | Large selection of text, image, audio, and realtime models. · Mature SDKs, tools, and structured outputs. · Usage-based billing suits variable workloads. | Strong models for coding and long documents. · Prompt caching and batch processing can reduce costs. · Clear developer documentation and tool-use capabilities. | A free tier eases prototyping and small projects. · Strong handling of text, image, audio, and video. · Good integration with the Google Cloud ecosystem. | Hundreds of models and many providers behind one API. · Pay-as-you-go simplifies model testing without many separate accounts. · Routing and fallbacks increase operational flexibility. | EU data centers are attractive for European projects. · Free and Scale modes support both entry and production use. · Text, code, OCR, and agent features from one platform. | Very deep macOS integration and fast keyboard workflows. · Multiple AI models, extensions, and BYOK in one tool. · Official time windows make many allowances understandable. | Agent, codebase context, and editor work directly together. · Supports macOS, Windows, and Linux. · Multiple models and tiered usage for different intensity levels. | BYOK and many models reduce vendor lock-in. · The gateway states it adds no markup to model prices. · Free models and open coding workflows make it easy to start. | More than 260 apps cover many productivity needs. · AI credits are transparently tiered monthly by plan. · Especially attractive when several included apps are already useful. | Local processing enables strong privacy without mandatory cloud use. · One-time purchase instead of a subscription for unlimited local transcription. · Many export formats and good integration into Mac workflows. | Many providers and local models in one native Mac app. · API keys are stored in Keychain and are not routed through a middle server. · One-time purchase options can be attractive for long-term use. | Supports VS Code, Visual Studio, Xcode, JetBrains, Neovim, and more. · Paid tiers combine unlimited completions with transparent AI credits. · Cloud agent, code review, CLI, and GitHub are tightly integrated. | Very strong image aesthetics and consistent creative results. · Relax Mode enables unlimited image generation from Standard upward. · Fast GPU time, concurrency, and add-on costs are documented concretely. | Deep integration with Express, Photoshop, and other Creative Cloud apps. · Unlimited standard image features remain available after premium credits are spent. · Firefly models use licensed and public-domain training sources plus Content Credentials. | Concrete credit costs per model, second, image, or audio unit. · All tiers allow commercial use of generated content. · Broad selection of native and third-party video and image models. |
| FitLimitation | Exact message and tool limits are dynamic and often only visible in-product. · Cloud processing remains required for core features. | Five-hour and weekly limits can constrain intensive use. · Fewer native image and media tools than some all-rounders. | Some Ultra features are region-restricted. · Concrete limits can change and partly depend on model and feature. | Many allowances are described as dynamic weekly classes instead of fixed counts. · API usage is billed separately from the consumer subscription. | Point consumption varies significantly by bot and model. · Purchase options and availability can vary by country and platform. | Exact public message limits and reset times are unavailable. · Plan access and features partly depend on X offerings. | Costs and rate limits depend on model and usage tier. · No ChatGPT subscription is included; interface and operations must be built separately. | No native image generation and a smaller media portfolio. · Rate limits and costs scale with model and usage tier. | Quotas differ by model, project, and billing tier. · Daily limits reset at midnight Pacific time. | Purchase fees and model-specific prices complicate simple cost comparisons. · Privacy and availability also depend on the selected upstream provider. | Exact organization limits are mostly visible only in the console. · Smaller model and media selection than large aggregators. | Most of the value is concentrated on macOS. · Premium models have additional hourly, daily, or weekly limits. | Included model usage is stated only approximately. · Intensive use can result in additional usage-based costs. | Availability, privacy, and limits partly depend on the upstream model. · Paid gateway use stops when the balance is empty. | Unused AI credits expire at the monthly reset. · Platforms and feature scope differ by app and plan. | Available only for macOS. · Some models and features require Apple Silicon or optional cloud services. | Provider API costs are additional to the app license. · macOS only; local model performance depends on the user’s Mac. | Agents and frontier models consume a model-dependent credit budget. · Individual-plan interactions may be used for model improvement unless users opt out. | Private generation through Stealth Mode starts only with Pro. · Companies above one million US dollars in revenue need Pro or Mega. | Partner models and premium media features can consume credits quickly. · Credits reset monthly and do not roll over. | Standard and Pro credits expire at the next billing date. · High-quality video can consume the monthly allowance very quickly. |
| Platforms and ecosystemsOperating system | Web, macOS, Windows, iOS, Android | Web, macOS, Windows, iOS, Android | Web, Android, iOS | Web, desktop browser, iOS, Android | Web, iOS, Android | X and Grok apps | REST API and official SDKs | Messages API and official SDKs | Gemini API and Google AI Studio | OpenAI-compatible API and web chat | Mistral API and Studio | macOS | macOS, Windows, Linux | VS Code extension and CLI | Mac, iPhone, iPad, and web depending on app and plan | macOS 14 or newer; Intel and Apple silicon | macOS 13 or newer; Intel and Apple silicon | GitHub, VS Code, Visual Studio, Xcode, JetBrains IDEs, Neovim, Eclipse, Raycast, Zed | Web and Discord | Web, mobile, Adobe Express, Photoshop, and compatible Creative Cloud apps | Web app |
| Platforms and ecosystemsEcosystem | Codex, Deep Research, files, images, voice | Claude Code, Cowork, projects, connectors, MCP | Google apps, NotebookLM, storage, Flow | Advanced models, Research, Create files and apps, Comet | Multiple text, image, video, and audio bots | Web and X search, files, images, code execution | Responses, tools, Batch, fine-tuning, realtime | Prompt caching, Batch, tools, MCP connector | Multimodal models, caching, Batch, files | 400+ models and 70+ providers on paid plans | Text, code, audio, OCR, agents | Launcher, extensions, AI, sync, BYOK | Editor, agents, Tab, MCP, cloud agents | Gateway credits, BYOK, multiple providers, MCP | 260+ apps and AI apps | Whisper, Parakeet, local models, optional cloud and automation integrations | OpenAI, Anthropic, Gemini, OpenRouter, Mistral, local LLMs, and more | GitHub, editors, CLI, code review, cloud agents, MCP | Not publicly specified | Adobe Firefly, Express, Photoshop, Illustrator, Premiere, Lightroom, and partner models | Runway and partner models for video, image, and audio generation |
| Privacy and restrictionsPrivacy | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Not publicly specified | Local transcription stays on-device; optional cloud services are separate. | Not publicly specified | Individual subscribers can opt out of model improvement in Copilot settings. | Not publicly specified | Not publicly specified | Not publicly specified |
| Privacy and restrictionsRestrictions and regions | Add-on credits currently apply only to selected supported features. | Claude surfaces share the same usage allowance. | Some Ultra features are available only in the United States. | The consumer subscription includes no complimentary API credits. | Purchase availability differs by country and platform. | Access and allowance depend on the selected X or SuperGrok plan. | API usage is billed separately from ChatGPT subscriptions. | The API is metered and billed in USD. | Higher tiers depend on billing, spend, and account age. | Credit purchases carry a fee; taxes are excluded. | Free mode is limited; Scale is pay as you go. | Raycast allowances do not apply to BYOK requests. | After exceeding the allowance, extra usage or a higher plan is required. | Paid requests stop at zero balance; provider limits may also apply. | Some apps and cloud features have Setapp-specific limits. | Not publicly specified | Model access requires a personal API key and separate provider charges. | Not publicly specified | Stealth Mode and use by companies above $1M annual gross revenue require Pro or Mega. | Premium video, audio, translation, and partner-model outputs consume credits. | Pricing is per editor while members share the workspace credit allowance. |
| EvidenceOfficial source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source | Official source |
| EvidenceFreshness | Outdated | Verified | Verified | Outdated | Needs review | Outdated | Needs review | Outdated | Verified | Verified | Needs review | Verified | Needs review | Verified | Verified | Verified | Verified | Verified | Outdated | Outdated | Verified |