Claude AI’s pricing structure includes free and paid plans for its web application, as well as usage-based pricing for its API. All pricing information below is provided directly by Anthropic and is current as of 2025. (This overview excludes any third-party platform pricing on services like AWS Bedrock or Google Cloud Vertex AI, focusing only on Anthropic’s own pricing.)
Claude.ai Plans for Individual Users
Anthropic offers several Claude.ai plans for individual users, ranging from a free tier to paid subscriptions with higher usage limits and additional features:
- Free Plan – $0 per month. The Free tier is available to everyone at no cost. It provides access to Claude’s core capabilities – users can chat with Claude on the web or mobile apps, generate content and code, analyze text or images, and even use web search within Claude. The Free plan comes with limited daily usage allowances (sufficient for basic use) and includes all fundamental features of the Claude assistant.
- Claude Pro – $20 per month (or approx. $17/month with an annual subscription, billed $200 upfront). The Pro plan is designed for everyday productivity and offers increased usage limits compared to Free (i.e. a higher quota of messages/tokens). Pro users get all Free features plus several enhancements: access to Claude Code (a coding assistant mode available on web and in terminal), the ability to create and execute code files, unlimited project workspaces to organize chats/documents, and integrations such as connecting Claude to Google Workspace tools. Pro subscribers also benefit from “extended thinking” mode for complex tasks and can utilize more Claude model variants than the free tier. This plan is ideal for professionals who regularly use Claude and need greater capacity and advanced features.
- Claude Max – $100 to $200 per month, depending on usage level. The Max plan is meant for power-users who need significantly higher throughput. It offers 5× the usage of Pro for $100/month, or 20× the usage of Pro for $200/month. Subscribers can choose the level that fits their needs (“Expanded Usage” at 5× or “Maximum Flexibility” at 20×). The Max plan includes everything in Pro, and further boosts the limits on output length and conversation memory. Max users also receive priority access during peak traffic and early access to new Claude features and models. This plan ensures minimal disruption for those working on demanding projects by providing substantially more Claude usage capacity.
Team and Enterprise Plans (Organizations)
For organizations and collaborative teams, Claude for Work is available in Team and Enterprise editions with per-seat pricing and additional enterprise features:
- Team Plan – Designed for businesses or groups (minimum 5 users required). The Team plan is priced per “seat” with two tiers: Standard seats at $25 per user/month (with annual subscription, or $30 if month-to-month), and Premium seats at $150 per user/month. Standard seats include core Claude chat, projects, and all Pro-level capabilities for each user. Premium seats (at a higher price) add Claude Code access for those users, enabling the integrated coding assistant features. All Team plans provide centralized admin controls (for managing connectors and permissions), enterprise-level integrations (e.g. connectors for Microsoft 365, Slack, etc.), an organization-wide search of chats/documents, unified billing, and early access to new collaboration features. Team plan pricing is billed per user and allows companies to provision Claude access with administrative governance.
- Enterprise Plan – Tailored for large-scale business deployments, with custom pricing available through Anthropic’s sales team. The Enterprise plan includes all Team features and goes further to meet enterprise requirements. Notably, it allows higher usage quotas (“More usage*” beyond even Team limits) and an enhanced context window for Claude’s models – meaning enterprise users can work with longer inputs or documents than standard plans support. Additionally, Enterprise offers advanced security and integration capabilities: Single Sign-On (SSO) and domain-based access control, SCIM user provisioning, audit logs of Claude usage, compliance APIs for monitoring, and custom data retention policies. Enterprise customers also retain access to Claude Code for technical users (included with premium seats) and receive dedicated support. Pricing is determined case-by-case based on the scale and needs of the organization (since usage can vary greatly at enterprise scale). Interested businesses should contact Anthropic for a quote.
- Education Plan – Anthropic provides a specialized plan for educational institutions (universities and schools). This offers comprehensive campus-wide access to Claude at discounted academic rates, covering students, faculty, and staff. The Education plan typically includes dedicated API credits for research, an academic “learning mode,” and training resources to support adoption. Universities can work with Anthropic to implement Claude AI broadly in curricula and research, benefiting from volume pricing and educational features. (This plan is available via Anthropic’s sales team and is not a standard commercial offering, but it is publicly mentioned for eligible institutions.)
Claude API Usage Pricing
For developers and enterprises integrating Claude via the Anthropic API, pricing is usage-based. There are no fixed monthly fees to use the API; instead, you pay per token processed by Claude. Charges are broken out by input tokens (the text prompts or data you send into Claude) and output tokens (the text generated by Claude in response). Prices vary by model variant, reflecting their relative capabilities and computational cost. All prices below are in USD and are per 1 million tokens (MTok):
- Claude Haiku – $1.00 per million input tokens and $5.00 per million output tokens. Haiku is the fastest and most cost-efficient model in the Claude 3 family. It offers solid general performance with an emphasis on speed and lower cost, making it suitable for high-volume or real-time applications where low latency and cost are priorities. (For example, processing 1,000 tokens of input with Haiku costs only about $0.001, and generating 1,000 tokens of output costs about $0.005 under this model’s rates.)
- Claude Sonnet – $3.00 per million input tokens and $15.00 per million output tokens (standard context). Sonnet is a general-purpose, highly intelligent model – Anthropic’s description is that it’s the “most intelligent” Claude model for complex reasoning, agents, and coding tasks. It provides a balance of advanced capability and faster runtime, at a moderate price point. The rates quoted ($3/$15) apply for prompts up to 200K tokens in length. Extended context: Sonnet can handle extremely long inputs beyond 200K tokens (up to ~1 million tokens in a special beta mode), but requests exceeding 200K tokens are billed at higher “long context” rates. In practice, this means if you utilize Sonnet’s extended context window, the cost doubles for input tokens (~$6.00/MTok) and increases by 50% for output tokens (~$22.50/MTok) beyond the 200K-token threshold. These higher rates reflect the additional computational load of very large context processing. For typical usage under 200K tokens, however, Sonnet remains at the base $3/$15 per million token pricing.
- Claude Opus – $15.00 per million input tokens and $75.00 per million output tokens. Opus is the most powerful and advanced model in Anthropic’s Claude 3 lineup, offering state-of-the-art reasoning and accuracy on complex tasks. It has the highest capacity and is priced accordingly at the top end. Opus is ideal for specialized workloads that demand the greatest level of model capability (for instance, intricate analyses, deep knowledge reasoning, or heavy creative tasks), but users should budget for the significantly higher token costs relative to Haiku or Sonnet. All usage of Opus (up to its standard context limit) is charged at the $15/$75 per million rate; there is no separate extended context pricing tier publicly indicated for Opus, which supports up to 200K tokens context by default.
Context Windows and Token Limits: All Claude 3 models accept very large context windows compared to many AI models. By default, each model can handle up to 200,000 tokens of input context in a single request (roughly equivalent to around 150,000 words of text). This allows Claude to ingest lengthy documents or conversations. For even longer inputs, Claude Sonnet has an extended context mode (beta) supporting up to 1,000,000 tokens in the prompt. When using such an extremely long context, the above-mentioned long context pricing for Sonnet applies beyond 200K tokens. Enterprise customers or specialized use cases can arrange access to these expanded context capabilities (Anthropic notes that inputs beyond 200K may be made available to select customers who need that scale).
In terms of output size, Claude can also generate very lengthy responses if needed: Claude Haiku and Sonnet can each produce up to 64K tokens of output, while Claude Opus outputs up to 32K tokens maximum in a single completion. These limits ensure reliability in generation. The vast majority of applications will not hit these upper bounds, but they are noteworthy for planning extremely large queries or documents.
Additional API Usage Fees: The Claude API ecosystem includes optional features and service tiers that carry their own pricing, as part of Anthropic’s direct offering:
- Prompt Caching: Anthropic offers a prompt caching mechanism to reduce costs on repetitive queries. Developers can have Claude “write” prompts/results to a cache and later “read” them back instead of regenerating. Using the cache is significantly cheaper than re-processing the same text. For example, reading cached content costs only $0.10 per million tokens with Claude Haiku (and similarly low rates for other models), whereas writing to the cache costs around 25% of the normal input token price. This feature helps applications that repeatedly use the same context or data to save on token usage bills.
- Web Search Tool: If you utilize Claude’s integrated web browsing/search capability via the API (giving Claude the ability to fetch live information from the web), there is a nominal fee for those external search requests. The cost is $10 per 1,000 web searches executed through Claude. (These search charges are in addition to the normal token costs for the content Claude processes as a result of the search.) This pricing applies to the “Web search” tool usage within Claude’s API.
- Code Execution Tool: Claude can run Python code in a sandbox (useful for data analysis, calculations, or other tool-assisted tasks). Anthropic includes a generous free allowance of 50 hours of sandbox runtime per day for each organization. If code execution needs exceed that free daily allotment, additional usage is billed at $0.05 per hour per container (i.e. per hour of compute time in the Claude code sandbox). This ensures that occasional code runs are free, but heavy automated use of Claude’s Python tool incurs a small incremental cost.
- Service Tiers (Priority vs. Standard vs. Batch): By default, API calls use the Standard service tier, which balances performance and cost. Anthropic also provides a Batch tier for asynchronous or high-throughput workloads that can be processed with more relaxed latency – batch processing jobs are charged at roughly 50% of the standard token rates, offering substantial savings for non-real-time tasks. Conversely, for mission-critical or time-sensitive deployments, a Priority tier is available, which ensures the highest availability and fastest response, albeit typically involving a custom arrangement (and potentially higher cost) through Anthropic’s sales team. Standard tier pricing is published (as above), while Batch tier pricing can be inferred as approximately half of standard (Anthropic explicitly notes “Save 50% with batch processing”), and Priority tier pricing is not publicly listed (organizations must discuss their requirements with Anthropic). These tiers allow customers to choose between optimal cost efficiency or guaranteed performance, depending on their use case.
All the above pricing is directly from Anthropic and subject to change with notice. Volume discounts or custom contracts may be available for large commitments (especially at enterprise scale), but the figures given reflect the publicly available rates and plan prices as of the latest update. Prospective users should evaluate their expected usage (in tokens or team size) against these prices to estimate costs. Anthropic’s pricing is designed to be transparent and scalable, from free personal use up to enterprise and API integration at scale, enabling organizations to adopt Claude AI according to their needs and budget.