Reviewed July 29, 2026. Anthropic’s live plan page and Claude Platform pricing documentation were checked for this refresh. Prices are in US dollars before tax and can change.
Verdict: Claude pricing is easier to understand once subscriptions and API usage are separated. Individuals pay for an access tier with usage limits; developers pay by tokens; the self-serve Enterprise plan combines a seat charge with API-rate usage. Pro is the sensible default for one regular user. Max only makes sense after Pro limits become a recurring cost, while API buyers should choose the cheapest model that meets the task rather than defaulting to the flagship.
Claude Plans at a Glance
| Plan | Current price | Best fit |
|---|---|---|
| Free | $0 | Occasional chat, files, web search and testing the workflow. |
| Pro | $20 monthly, or $17/month with $200 billed annually | One frequent user who needs more usage, Claude Code, projects and research. |
| Max | From $100/month | Heavy individual use; choose 5× or 20× the Pro usage level. |
| Team standard | $25/seat monthly, or $20/seat/month annually | Teams needing shared administration, SSO and no training on content by default. |
| Team premium | $125/seat monthly, or $100/seat/month annually | Heavy team users needing 5× standard-seat usage. |
| Enterprise self-serve | $20/seat plus usage at API rates | Organizations that need spend controls, audit features and usage-based billing. |
“More usage” is not the same as unlimited use. Limits vary with model, message length, attachments and product features. Buy Max because measured interruptions justify it, not because the headline sounds safer.
Claude API Pricing
| Model | Input per million tokens | Output per million tokens |
|---|---|---|
| Claude Fable 5 | $10 | $50 |
| Claude Opus 5 | $5 | $25 |
| Claude Sonnet 5 | $2 through August 31, then $3 | $10 through August 31, then $15 |
| Claude Haiku 4.5 | $1 | $5 |
Those are base token rates. Prompt caching, long-context requests, batch processing, fast mode, regional inference and tool use can change the total. Anthropic currently prices five-minute cache reads below ordinary input, so repeated system prompts or reference material can be cheaper when caching is designed deliberately.
How to Estimate a Real API Bill
For a simple request, multiply input tokens by the input rate and output tokens by the output rate, then divide each by one million. A workload with 800,000 Sonnet 5 input tokens and 200,000 output tokens during the introductory period is about $3.60 before caching or tools: $1.60 for input and $2 for output.
- Count retries. Failed or rejected application-level outputs can still consume billable tokens.
- Watch output length. Output costs more than input on every current model in the table.
- Route by task. Classification and extraction may not need the model used for difficult coding or agent work.
- Set spend limits. A per-user subscription limit and an API budget are separate controls.
Which Claude Plan Should You Choose?
Start with Free if your use is sporadic, Pro if Claude is a daily work tool, and Max only after tracking how often Pro interrupts valuable work. Team is about administration as much as volume. Enterprise is the notable 2026 pricing change because its seat-plus-usage structure shifts cost toward actual consumption.
Developers should prototype with Sonnet or Haiku, measure quality and cost on a representative set, then escalate only the hard cases. For model-specific context, compare Kingy’s Claude Opus 5 guide and Claude Fable 5 analysis.
Official Sources
The Kingy Brief
Follow The Kingy Brief.
One consequential launch, one pricing, limit, or shutdown change, one hands-on test, one exact prompt or Test Pack, and one try / watch / skip verdict.
Free · Choose your subjects · Double opt-in · Unsubscribe anytime