Claude Plans And Pricing

The Pricing Split

Claude has two pricing worlds. The first is subscription access to Claude products: Free, Pro, Max, Team, and Enterprise. The second is API usage through the Claude Platform and cloud partners, where billing is based on model, input tokens, output tokens, caching, search, code execution, batches, and service features.

This split is the source of many confused comparisons. A Pro or Max subscription can include Claude Code access, but that is not the same as free API usage through the Console. Conversely, API billing is useful for production apps and batch workloads even when a person also has a paid Claude subscription.

Individual Plans

The Choose a Claude plan support page describes four individual tiers: Free for occasional use, Pro for regular use, and Max at two usage multiples for frequent and daily users. Usage capacity is described in relative terms rather than as a public fixed message count, and it depends on message length, attached files, conversation length, model, and feature usage.

Claude Central does not republish Pro or Max subscription prices, because subscription pricing changes without a documented announcement more often than API pricing does. Check the official pricing page and the Help Center plan article, and verify on Anthropic before budgeting. What is documented and stable is the model access story: Opus 5 is the default model on Claude Max and the strongest model available on Claude Pro.

Teachers And Researcher Tracks

Claude for Teachers, announced July 14, 2026, gives verified US K-12 educators free premium access along with teaching skills and curricula mapped to all 50 states. Sign-up runs on a one-year window and a school and district offering launched August 28, 2026. The program is FERPA-configured, and identifiable student records require school or district authorization.

The August 27, 2026 scientist program allocates 10,000 seats: a standard tier free and a premium tier at $15 per month for five times the usage, for one year. Biology and chemistry researchers are limited to Opus-class models, and Fable still blocks professional biology and drug-development work. Claude Science remains the research workbench.

Team And Enterprise

Team is not just “Pro with colleagues.” It introduces organizational billing, seat management, collaboration, and per-member usage limits. The Team support article describes Standard and Premium seats, a minimum team size, and a seat cap before upgrading to Enterprise. Verify the current seat prices and thresholds on the Help Center article, which is the owner of that fact.

Enterprise is the governance tier. The Enterprise support article says it includes Team features plus security and compliance controls such as audit logs, SCIM, custom data retention, Compliance API, Analytics API, and customer-managed encryption keys. Enterprise pricing works differently: seat fees cover access and usage is billed separately at API rates.

API Token Pricing

API prices vary by model and token direction. The figures below are per million tokens, input first and output second, from the Platform pricing page as checked on September 19, 2026.

The single most important pricing change this quarter is that Sonnet 5’s introductory $2/$10 rate became the standard rate. The increase to $3/$15 that was scheduled for September 1, 2026 did not take effect.

• Claude Fable 5.1 — $10 input / $50 output — cache reads $0.25 (2.5% of base input).
• Claude Opus 5 — $5 input / $25 output — cache reads 10% of base input. Fast mode runs roughly 2.5x speed at 2x base price.
• Claude Sonnet 5 — $2 input / $10 output — now standard, not introductory.
• Claude Haiku 4.5 — $1 input / $5 output.
• Claude Mythos 5.1 — $10 input / $50 output with $0.25 cache reads, trusted access only.
• Batch API — 50% off for asynchronous workloads on supported models.

Modeling Real Cost

List price is a poor predictor of bill size. The levers that actually move the number are prompt caching, batch processing, web search and code execution charges, data residency, service tiers, and cloud platform pricing. Before budgeting, model a real workload with expected input, output, retry, cache-hit, and batch rates.

Fable 5.1’s 2.5% cache reads are the standout: Anthropic reports roughly 25% savings on typical workloads and up to about 45% on highly agentic ones versus Fable 5. That can make Fable 5.1 competitive with Opus 5 on cache-heavy agent loops even though its list price is double.