Updated 8 Jun 2026 • 7 mins read
Claude Pricing 2026: Complete Guide to Plans, API Costs & Real Token Rates
Claude offers free, Pro ($20/mo), Max ($100–$200/mo), and Team plans for chat users, plus a pay-as-you-go API: Haiku at $1/$5, Sonnet at $3/$15, and Opus at $5/$25 per million tokens. Batch and caching discounts can cut costs by up to 95%.
If you are budgeting for Claude in 2026, the published price list only tells half the story. Claude is sold two completely different ways, and confusing them is the most common reason teams overspend. On one side are flat monthly subscriptions on claude.ai for people who chat with Claude in a browser or app. On the other is pay-as-you-go API pricing measured in tokens, billed per million, for developers who build Claude into their own products. In this guide we break down every Claude pricing tier in 2026, from the Free plan to the $200 Max tier to per-token API rates for Opus, Sonnet, and Haiku, plus the levers that quietly decide your real bill: prompt caching, batch processing, the 1M-token context window, and tokenizer changes that raise effective cost even when the headline rate stays flat.
Key Statistics: Claude Pricing in 2026 at a Glance
The numbers below reflect Anthropic's public pricing as observed in mid-2026. Figures are USD and API rates are per million tokens (MTok).
| Metric | Applies to | 2026 Figure |
|---|---|---|
| Cheapest current API model | Claude Haiku 4.5 | $1 in / $5 out per MTok |
| Balanced workhorse model | Claude Sonnet 4.6 | $3 in / $15 out per MTok |
| Flagship model (current) | Claude Opus 4.8 | $5 in / $25 out per MTok |
| Output-to-input price ratio | All current models | Output costs 5× input |
| Batch API discount | All models | 50% off every token |
| Prompt caching savings | Cached input | Up to 90% cheaper |
| Long context | Opus 4.8/4.7/4.6, Sonnet 4.6 | 1M tokens at standard rate |
| Entry paid subscription | Claude Pro | $20/mo ($17/mo annual) |
| Top consumer subscription | Claude Max 20x | $200/mo (20× Pro usage) |
| Team plan minimum | Claude Team | 5 seats, from $25/seat/mo |
How Claude Pricing Actually Works
Subscriptions buy capacity, not tokens: you pay a fixed monthly fee for a usage budget, with heavier plans unlocking more messages, more Opus access, and priority at peak load. The API is the opposite, with no monthly fee and no included usage, you pay only for tokens sent and received. The break-even point moves fast: a solo analyst living in the chat app is almost always better off on a subscription, while a product team piping Claude through the Messages API at scale needs token math. For the full feature picture of the chat product, our complete Claude AI 2026 guide covers models, modes, and hidden commands.
Claude Subscription Pricing (claude.ai)
Subscriptions are the right fit for individuals and teams who use Claude through the web, desktop, or mobile app rather than through code.
| Plan | 2026 Price | What you get |
|---|---|---|
| Free | $0 | Sonnet 4.6 default, limited Haiku, web search, file uploads, Artifacts, daily limits |
| Pro | $20/mo ($17 annual) | 5× Free capacity, all models incl. Opus, Claude Code, projects, integrations |
| Max 5x | $100/mo | 5× Pro usage, priority access, early features |
| Max 20x | $200/mo | 20× Pro usage, top priority routing, heavy Opus headroom |
| Team | from $25/seat/mo (min 5) | Shared workspace, admin controls; Premium seat adds Claude Code |
| Enterprise | Custom | SSO, compliance, dedicated support, SLAs, tailored integrations |
Claude API Pricing (per million tokens)
API pricing is where engineering and FinOps teams need precision. Rates are per million tokens, billed separately for input and output, with output consistently five times the input rate across the current lineup.
| Model | Input /MTok | Output /MTok | Best for |
|---|---|---|---|
| Claude Haiku 4.5 | $1.00 | $5.00 | High-volume routing, extraction, classification |
| Claude Sonnet 4.6 | $3.00 | $15.00 | Best price-to-quality balance for most production work |
| Claude Opus 4.6 | $5.00 | $25.00 | Flagship-class reasoning |
| Claude Opus 4.7 | $5.00 | $25.00 | Flagship; note new tokenizer can add ~35% tokens vs 4.6 |
| Claude Opus 4.8 | $5.00 | $25.00 | Current flagship (launched May 2026); Fast Mode $10/$50 |
| Legacy Opus 4.1 | $15.00 | $75.00 | 3× pricier; pin only for compatibility |
| Legacy Haiku 3 | $0.25 | $1.25 | Cheapest legacy option for simple tasks |
Two competitor guides are worth reading alongside this table to sanity-check value: our ChatGPT pricing in 2026 breakdown and our Google Gemini API pricing guide. If you consume Claude through AWS, the rate structure differs again, which we cover in our Amazon Bedrock pricing explainer.
The Cost Levers That Decide Your Real Bill
Caching stores frequently reused context, such as a long system prompt or reference document, so repeated requests read it at up to 90% off the input rate, the largest single saving for RAG systems and agents. The Batch API processes requests within 24 hours at a flat 50% discount on every token, halving the bill for any job that does not need a real-time answer. Opus 4.8, 4.7, 4.6 and Sonnet 4.6 all support a 1M-token context at standard rates with no surcharge, though a bigger window still means more input tokens per call.
The tokenizer trap
Opus 4.7 introduced a tokenizer that can generate up to 35% more tokens for the same text than Opus 4.6. The per-token rate is unchanged, but a team migrating from 4.6 can see effective cost per request rise by up to a third. The 4.7 to 4.8 jump carries no further tokenizer penalty.
Falling headline rates can hide rising bills. As we explain in our piece on why falling token prices don't lower costs, usage growth almost always outpaces per-token cuts, which is why governance matters more than chasing the lowest sticker rate.
Subscription vs API: Which Is Cheaper?
This is the question most teams get wrong. The comparison table below frames the decision by usage pattern rather than by feature list.
| Usage pattern | Cheaper path | Why |
|---|---|---|
| Light personal chat use | Subscription (Pro) | Flat $20 beats per-token unpredictability for daily writing and research |
| Heavy interactive use | Subscription (Max) | Predictable cost; priority access during peak; no metering anxiety |
| Small product, low volume | API (Haiku/Sonnet) | Pay only for what ships; no idle seat cost |
| High-volume production | API + caching + batch | Token controls cut 50–95% versus naive usage |
| Mixed org (chat + build) | Team/Enterprise + API | Seats for people, API for products, billed separately |
Expert Insights: Controlling Claude Spend in Practice
In our work helping FinOps and engineering teams allocate AI spend, three patterns separate teams that stay on budget from those that get surprised.
- Pick the model family first. The single biggest lever is choosing the cheapest model that clears your quality bar, then optimizing within it. Routing simple calls to Haiku and reserving Opus for genuinely hard reasoning often cuts spend more than any micro-optimization.
- Treat tokens like a metered utility. Set per-feature and per-team token budgets, alert on anomalies, and review cost-per-outcome rather than total spend.
- Tag and allocate from day one. AI bills behave like cloud bills, and the same FinOps discipline applies. Our guide to FinOps for AI token and GPU costs and our token budgeting framework lay out exactly how to instrument this.
For broader competitive and safety context before you commit, our Anthropic vs OpenAI enterprise comparison weighs models, pricing, and posture side by side.
FAQs
Claude is free to start. Paid consumer plans are Pro at $20/month, Max at $100 or $200/month, and Team from $25 per seat per month. API access is pay-as-you-go: $1/$5 per million tokens for Haiku 4.5, $3/$15 for Sonnet 4.6, and $5/$25 for Opus 4.8.
Claude Haiku 4.5 at $1 input and $5 output per million tokens is the cheapest current-generation model. Legacy Haiku 3 is even lower at $0.25/$1.25 for very simple tasks.
Claude Opus 4.8 costs $5 per million input tokens and $25 per million output tokens, the same as Opus 4.7 and 4.6. Its Fast Mode is $10/$50 per million tokens.
No. The API is pay-as-you-go with no monthly fee. You pay only for tokens sent and received. This is separate from claude.ai subscriptions like Pro at $20/month.
Claude Pro is $20/month ($17 billed annually) and gives about 5x the Free capacity, access to all models including Opus, plus Claude Code and integrations. It suits professionals who use Claude daily.
Claude Max is $100/month for 5x Pro usage or $200/month for 20x Pro usage. Max plans are monthly-only in 2026 with no annual discount.
Use prompt caching for up to 90% off repeated context, batch non-urgent requests for a flat 50% discount, and route simple tasks to Haiku while reserving Opus for hard problems.
Opus 4.8, 4.7, 4.6 and Sonnet 4.6 support a 1M-token context at standard rates with no surcharge, though larger inputs naturally consume more tokens per call.
Claude Sonnet 4.6 at $3/$15 sits above some competitor mid-tier models, while Claude's premium tier holds at $5/$25. The right choice depends on quality per task, not sticker rate alone.
Yes. The Free plan costs nothing, defaults to Sonnet 4.6 with limited Haiku, and includes web search, file uploads, and Artifacts subject to daily limits. The API has no permanent free tier.