Updated October 2026
Claude pricing has changed more in 2026 than in any year since Anthropic launched its API. A new frontier tier arrived, the flagship got cheaper, a planned Sonnet price increase was cancelled, and the subscription plans now stretch from free to $200 a month.
This guide covers Claude pricing as it stands in October 2026: every subscription plan, every current API rate, what the hidden multipliers are, and worked examples that show what a real workload costs on each model. All prices come from Anthropic’s API pricing page and Claude plans page, which you should check before committing budget, since rates do change.
Claude Pricing at a Glance
Here’s Claude pricing 2026 at a glance, if you only need the numbers.
Subscription plans (per month):
| Plan | Monthly billing | Annual billing | Who it’s for |
|---|---|---|---|
| Free | $0 | $0 | Trying Claude |
| Pro | $20 | $17 | Individuals, includes Claude Code |
| Max 5x | $100 | n/a | Heavy individual users |
| Max 20x | $200 | n/a | Power users and developers |
| Team Standard | $25 per seat | $20 per seat | Small teams |
| Team Premium | $125 per seat | $100 per seat | Teams that need higher limits |
| Enterprise | Custom | Annual contract | Larger organizations |
API rates (per million tokens):
| Model | Input | Output | Cache read |
|---|---|---|---|
| Claude Fable 5.1 | $10.00 | $50.00 | $0.25 |
| Claude Opus 5.5 | $4.00 | $20.00 | $0.20 |
| Claude Sonnet 5 | $2.00 | $10.00 | $0.20 |
| Claude Haiku 4.5 | $1.00 | $5.00 | $0.10 |
Plan details are summarized from Benchlm’s plan breakdown, and API rates from Benchlm’s API pricing tracker. Both are checked against Anthropic’s own pages.
The Claude Model Lineup in October 2026
Understanding the lineup is the foundation of managing Claude pricing, because the same task can cost ten times more on one model than another.
Claude Fable 5.1 is Anthropic’s Mythos-class frontier model, released on September 1. It costs $10/$50 per million tokens, the same as Fable 5, but with cache reads 75% cheaper, according to Anthropic’s Fable page. It’s built for the hardest reasoning and long autonomous tasks.
Claude Opus 5.5 arrived on September 22 at $4/$20, a 20% cut from Opus 5’s $5/$25. Anthropic positions it at Fable 5.1 level on most work, which makes it the flagship most teams should actually use. It has a 1 million token context window with no long-context surcharge, per eesel’s pricing analysis. Our Claude Opus 5.5 pricing guide covers it in depth, including four breaking API changes to check before upgrading.
Claude Sonnet 5 launched on June 30 at an introductory $2/$10. That price was due to rise to $3/$15 after August 31, but Anthropic cancelled the increase in August, so $2/$10 is now the standard rate. Sonnet is the practical default for most production work.
Claude Haiku 4.5 is the cheapest current model at $1/$5, suited to high-volume, simple tasks like classification and extraction.
Older models. Opus 5 remains available at $5/$25. Unless your prompts are tuned for it, Opus 5.5 does the same job for less. Anthropic has said Sonnet 5.5 and Haiku 5.5 are on the way, so expect the lower tiers to change again.
How Claude Pricing Changed in 2026
It helps to know how Claude pricing got here, because several articles online still quote outdated rates.
- June 9: Claude Fable 5 launched at $10/$50 per million tokens, Anthropic’s first Mythos-class model for general use. It was briefly suspended under US export controls and restored on July 1, as explained in Anthropic’s statement.
- June 30: Claude Sonnet 5 launched at an introductory $2/$10, with a planned rise to $3/$15 after August 31.
- August: Anthropic cancelled the Sonnet 5 increase, making $2/$10 permanent.
- September 1: Fable 5.1 replaced Fable 5 at the same price, with cache reads 75% cheaper.
- September 22: Opus 5.5 launched at $4/$20, down 20% from Opus 5.
The direction is clear: Anthropic has been cutting Claude pricing at the top and middle of its lineup rather than raising it. If an article tells you Sonnet 5 costs $3/$15, or that Opus 4.8 is Anthropic’s flagship, it was written before these changes.
How Claude API Pricing Works
Claude API pricing is per token, which is roughly three-quarters of an English word. You pay separately for input (what you send) and output (what Claude writes back).
Three rules explain most Claude pricing surprises:
- Output costs five times as much as input on every current model. Long answers and multi-step agents are where bills grow.
- The tiers are far apart. Fable 5.1 costs ten times as much as Haiku 4.5 for the same tokens.
- Repeated input can be almost free. Cache reads cost about a tenth of the normal input rate, or less.
Worked example: the same workload on every model
Assume a team sends 10 million input tokens and receives 2 million output tokens a month. That’s a modest production workload, such as an internal assistant used across a small company.
| Model | Monthly cost |
|---|---|
| Claude Fable 5.1 | $200 |
| Claude Opus 5 | $100 |
| Claude Opus 5.5 | $80 |
| Claude Sonnet 5 | $40 |
| Claude Haiku 4.5 | $20 |
This is the clearest way to see how Claude pricing scales across the lineup. The spread is the whole story of Claude pricing. Choosing the right tier matters more than any other optimization.
Note the Opus line too. On raw rates, Opus 5.5 is 20% cheaper than Opus 5 for identical token counts. Anthropic says it’s about 40% cheaper on typical workloads, because it finishes tasks in fewer tokens. That second figure depends on your work, so measure it rather than assuming it.
The Multipliers Behind Claude Pricing
The headline rates aren’t the whole Claude pricing story. Several options move the price up or down.
Prompt caching (cheaper). When the same content, like a long system prompt or document, is sent repeatedly, Claude can read it from cache. Writing to the cache costs 1.25 times the input rate for a 5-minute cache, or 2 times for a 1-hour cache. Reading from it costs a small fraction of the input rate. Caching pays off when content is reused many times.
Batch processing (cheaper). Jobs that can wait up to 24 hours cost 50% less on input and output.
Fast mode (more expensive). Opus 5.5’s Fast mode runs about 2.5 times faster at double the price: $8/$40 per million tokens, according to eesel.
US-only inference (more expensive). Requiring processing in the US adds a 1.1x multiplier, per the same source.
Worked example: caching and batch on Sonnet 5
Take the 10 million input and 2 million output workload on Sonnet 5, which costs $40 at standard rates.
If 70% of that input is repeated content served from cache, input drops from $20 to $7.40: 3 million tokens at $2.00 ($6) plus 7 million at $0.20 ($1.40). Output stays at $20, so the total falls from $40 to $27.40, a 31.5% saving. The small one-time cost of writing content to the cache is extra, but it’s minor when the content is reused many times.
If the same workload can run as an overnight batch instead, it costs $20, half the standard price.
Claude Pro vs API: Which Should You Choose?
This is the most common Claude pricing question, and the answer depends on how you use Claude.
Choose a subscription (Pro or Max) if you use Claude yourself through the app, the website or Claude Code. Pro at $20 a month includes Claude Code and access to Opus, Sonnet and Haiku, with Fable available through usage credits. Max at $100 or $200 raises your usage limits substantially and includes Fable within its limits.
Choose the API if you’re building Claude into a product, an automation or an internal tool, or if your usage is spiky and you’d rather pay only for what you use.
A note on value. Subscriptions describe limits as multiples of usage rather than token counts, so a direct comparison is hard. As a rough guide, $20 of API spend on Sonnet 5 covers about 660 tasks of 10,000 input and 1,000 output tokens each, at about $0.03 per task. If you use Claude heavily every day, a subscription is almost always better value. If you build software on top of it, the API is the only option.
Pro does not include API access. The two are billed separately, and API usage needs its own account on the Claude Platform.
Claude Pricing for Teams
For organizations, Claude pricing comes down to Team or Enterprise plans.
Team Standard costs $25 per seat per month, or $20 with annual billing. Team Premium costs $125 per seat per month, or $100 annually, for members who need much higher usage limits. Team plans cover groups of 2 to 150 people.
Enterprise pricing is custom, with an annual contract, a per-seat fee and usage billed at API rates. It adds administrative and security controls that larger companies need.
Worked example: a 10-person team
Ten people on Team Standard cost $2,400 a year with annual billing, or $3,000 a year on monthly billing. Moving the same ten people to Team Premium would cost $12,000 a year on annual billing. Most teams do best with Standard seats for everyone and Premium seats only for their heaviest users.
Claude Pricing vs ChatGPT
Claude pricing and OpenAI pricing now sit in strikingly similar bands.
| Tier | Claude | OpenAI |
|---|---|---|
| Consumer entry plan | Pro, $20/month | ChatGPT Plus, $20/month |
| Frontier API | Fable 5.1, $10/$50 | GPT-6 Astra, $10/$50 |
| Mid-tier API | Sonnet 5, $2/$10 | GPT-6 Sol, $2/$10 |
| Budget API | Haiku 4.5, $1/$5 | GPT-6 Luna, $0.10/$0.50 |
OpenAI prices are from Digital Applied’s GPT-6 coverage and OpenAI’s ChatGPT pricing page.
At the frontier and in the middle, the prices match exactly, so the decision comes down to quality on your tasks. Claude’s distinctive option is Opus 5.5 at $4/$20, which sits between the two and has no direct OpenAI equivalent. At the budget end, GPT-6 Luna is ten times cheaper than Haiku 4.5, which matters for very high-volume simple work.
For the full matchup, see our GPT-6 vs Claude Opus 5.5 comparison and GPT-6 pricing guide.
When to Use Each Claude Model
The single biggest lever in Claude pricing is picking the right model for each job.
Use Haiku 4.5 for high-volume simple work: classifying tickets, extracting fields from documents, tagging content and short routine replies.
Use Sonnet 5 for most business work: writing, standard coding, data analysis, customer support and research. It’s the default for good reason.
Use Opus 5.5 for complex reasoning, difficult coding, long documents and agentic tasks that Sonnet struggles with. At $4/$20, it’s affordable enough to use regularly.
Use Fable 5.1 only for the hardest problems, where you’ve tested Opus 5.5 and it falls short. At $10/$50, it should be a deliberate choice, not a default.
The smartest setup routes each task to the cheapest model that handles it well. Be careful with benchmark scores when choosing, since vendor-run results often look better than independent ones. Our guide on how to read AI benchmarks explains what to trust.
Hidden Costs in Claude Pricing
A few parts of Claude pricing catch teams off guard.
Agents multiply tokens. An agent that plans, calls tools and retries sends the model many requests per task, each carrying more context. A single agent task can cost dozens of times more than a chat answer. Our guide to the hidden cost of AI agents covers this in detail.
Caching only works if prompts stay identical. Integrations that rebuild prompts slightly differently on each request miss the cache discount entirely.
Model upgrades take work. Opus 5.5 came with breaking API changes, and prompts tuned for one model often need re-testing on the next. Budget developer time for upgrades.
Subscription seats add up. Seats for people who rarely use Claude are the easiest savings to find in an audit.
For the full picture across vendors, see our breakdown of the hidden cost of AI.
5 Ways to Lower Your Claude Bill
These five habits lower the Claude pricing you actually pay, without switching vendors.
- Default to Sonnet 5, move up to Opus 5.5 only when Sonnet falls short, and push simple tasks down to Haiku 4.5.
- Turn on prompt caching and keep repeated instructions identical at the start of each prompt.
- Batch anything that can wait. Overnight jobs cost half price.
- Cap agent steps and track cost per completed task, not per request.
- Audit subscriptions quarterly. Downgrade or cancel seats that aren’t being used.
For more practical savings, see our guide to reducing AI costs for small business and our AI pricing comparison.
The Bottom Line on Claude Pricing
Claude pricing in late 2026 is simpler than it looks. For individuals, Pro at $20 a month covers most needs, and Max is for people who hit Pro’s limits. For teams, Standard seats with a few Premium seats is the usual answer.
For developers, the API lineup has a clear structure: Haiku for volume, Sonnet as the default, Opus 5.5 for hard problems, and Fable 5.1 for the rare cases that need it. Add caching and batch processing, and most teams can cut their Claude bill substantially without touching quality.
FAQs About Claude Pricing
1. How much does Claude cost per month?
Claude has a free plan. Pro costs $20 a month ($17 with annual billing), Max costs $100 or $200 a month, Team seats cost $25 or $125 per person per month ($20 or $100 with annual billing), and Enterprise pricing is custom. API usage is billed separately, per token.
2. What is Claude API pricing per million tokens?
As of October 2026: Fable 5.1 costs $10 input and $50 output, Opus 5.5 costs $4 and $20, Sonnet 5 costs $2 and $10, and Haiku 4.5 costs $1 and $5. Batch processing halves these rates, and cached input costs a small fraction of the normal input price.
3. Is Claude Sonnet 5 still $2/$10?
Yes. Sonnet 5 launched at an introductory $2/$10 that was due to rise to $3/$15 after August 31, 2026. In August, Anthropic cancelled the increase and made $2/$10 the standard rate.
4. Is Claude cheaper than ChatGPT?
At most tiers they’re priced the same. Claude Pro and ChatGPT Plus both cost $20 a month, Fable 5.1 and GPT-6 Astra both cost $10/$50, and Sonnet 5 and GPT-6 Sol both cost $2/$10. Claude’s Opus 5.5 at $4/$20 has no direct OpenAI equivalent, while OpenAI’s GPT-6 Luna is much cheaper than Haiku 4.5 at the budget end.
5. Does Claude Pro include API access?
No. Pro covers the Claude apps, website and Claude Code. API usage requires a separate Claude Platform account and is billed per token.
6. Should I use Opus 5.5 or Fable 5.1?
Start with Opus 5.5. It costs 60% less than Fable 5.1 per token, and Anthropic positions it at Fable 5.1 level on most work. Move to Fable 5.1 only for tasks where you’ve tested Opus 5.5 and found it falls short.
7. How can I reduce my Claude API costs?
Match each task to the cheapest model that handles it well, use prompt caching for repeated content, run non-urgent work through the batch API at half price, and limit how many steps your agents can take. In our worked example, caching alone cut a Sonnet 5 bill by about 31%.
Prices reflect Anthropic’s pricing pages and linked sources as of October 2026. Always confirm current rates before making budget decisions.
Mahdi Ayadi writes and edits AI Empire Media. He covers AI models, tools and pricing with one question in mind: is this actually worth paying for?
Before writing about software, Mahdi spent more than seven years selling it, in B2B SaaS sales and account management across cybersecurity, marketing technology and travel technology. That experience shapes how he reads vendor pricing pages, benchmark claims and feature lists, and it’s why his articles focus on real costs, trade-offs and sources readers can check for themselves.
He also builds AI-powered automation for sales and onboarding workflows, which informs his coverage of AI agents and automation tools. He works in English, French, Arabic and Russian.